Real Time Machine Learning Based Car Detection in Images With Fast Training

Our primary interest is to build fast and reliable object recognizers in images based on small training sets. This is important in cases where the training set needs to be built mostly manually, as in the case that we studied, the recognition of the Honda Accord 2004 from rear views. We describe a novel variant of the AdaBoost based learning algorithm, which builds a strong classifier by incremental addition of weak classifiers (WCs) that minimize the combined error of the already selected WCs. Each WC is trained only once, and examples do not change their weights. We describe a set of appropriate feature types for the considered recognition problem, including a redness measure and dominant edge orientations. The existing edge orientation bin division was improved by shifting so that all horizontal (vertical, respectively) edges belong to the same bin. We propose to pre-eliminate features whose best threshold value is near the trivial position at the minimum or maximum of all threshold values. This is a novel method that has reduced the training set WC quantity to less than 10% of its original number, greatly speeding up training time, and showing no negative impact on the quality of the final classifier. We also modified the AdaBoost based learning machine. Our experiments indicate that the set of features used by Viola and Jones and others for face recognition was inefficient for our problem, recognizing cars accurately and in real time with fast training. Our training method has resulted in finding a very accurate classifier containing merely 30 WCs after about 1 h of training. Compared to existing literature, we have overall achieved the design of a real time object detection machine with the least number of examples, the least number of WCs, the fastest training time, and with competitive detection and false positive rates.     

An alignment-free fingerprint bio-cryptosystem based on modified Voronoi neighbor structures

Bio-cryptography is an emerging security technology which combines cryptography with biometrics. A good bio-cryptosystem is required to protect the privacy of the relevant biometric data as well as achieving high recognition accuracy. Fingerprints have been widely used in bio-cryptosystem design. However, fingerprint uncertainty caused by distortion and rotation during the image capturing process makes it difficult to achieve a high recognition rate in most bio-cryptographic systems. Moreover, most existing bio-cryptosystems rely on the accurate detection of singular points for fingerprint image pre-alignment, which is very hard to achieve, and the image rotation transformation during the alignment process can cause significant singular point deviation and minutiae changes. In this paper, by taking full advantage of local Voronoi neighbor structures (VNSs), e.g. local structural stability and distortion insensitivity, we propose an alignment-free bio-cryptosystem based on fixed-length bit-string representations extracted from modified VNSs, which are rotation- and translation-invariant and distortion robust. The proposed alignment-free bio-cryptosystem is able to provide strong security while achieving good recognition performance. Experimental results in comparison with most existing alignment-free bio-cryptosystems using the publicly-available databases show the validity of the proposed scheme.     

Optimal BSR Solutions to Several Convex Polygon Problems

This paper focuses on BSR (Broadcasting with Selective Reduction) implementation of algorithms solving basic convex polygon problems. More precisely, constant time solutions on a linear number, max(N, M) (where N and M are the number of edges of the two considered polygons), of processors for computing the maximum distance between two convex polygons, finding critical support lines of two convex polygons, computing the diameter, the width of a convex polygon and the vector sum of two convex polygons are described. These solutions are based on the merging slopes technique using one criterion BSR operations.     

Connectivity Based k-Hop Clustering in Wireless Networks

In this paper we describe several new clustering algorithms for nodes in a mobile ad hoc network. The main contribution is to generalize the cluster definition and formation algorithm so that a cluster contains all nodes that are at distance at most k hops from the clusterhead. We also describe algorithms for modifying cluster structure in the presence of topological changes. We also proposed an unified framework for most existing and new clustering algorithm where a properly defined weight at each node is the only difference in otherwise the same algorithm. This paper studied node connectivity and node ID as two particular weights, for k=1 and k=2. Finally, we propose a framework for generating random unit graphs with obstacles.     

A social network approach to trust management in VANETs

In the past few years, vehicular ad hoc networks(VANETs) was studied extensively by researchers. VANETs is a type of P2P network, though it has some distinct characters (fast moving, short lived connection etc.). In this paper, we present several limitations of current trust management schemes in VANETs and propose ways to counter them. We first review several trust management techniques in VANETs and argue that the ephemeral nature of VANETs render them useless in practical situations. We identify that the problem of information cascading and oversampling, which commonly arise in social networks, also adversely affects trust management schemes in VANETs. To the best of our knowledge, we are the first to introduce information cascading and oversampling to VANETs. We show that simple voting for decision making leads to oversampling and gives incorrect results in VANETs. To overcome this problem, we propose a novel voting scheme. In our scheme, each vehicle has different voting weight according to its distance from the event. The vehicle which is more closer to the event possesses higher weight. Simulations show that our proposed algorithm performs better than simple voting, increasing the correctness of voting.     

The influence of textual and verbal word-of-mouth on website usability and visual appeal

The Journal of Supercomputing - Word-of-Mouth (WOM) may impact the perception and experience of website usability and visual appeal. This study aimed to highlight the effects of WOM, implemented...     

Hop count optimal position-based packet routing algorithms for ad hoc wireless networks with a realistic physical Layer

Existing routing and broadcasting protocols for ad hoc networks assume an ideal physical layer model. We apply the log-normal shadow fading model to represent a realistic physical layer and use the probability p(x) for receiving a packet successfully as a function of distance x between two nodes. We define the transmission radius R as the distance at which p(R)=0.5. We propose a medium access control layer protocol, where receiver node acknowledges packet to sender node u times, where u*p(x)/spl ap/1. We derived an approximation for p(x) to reduce computation time. It can be used as the weight in the optimal shortest hop count routing scheme. We then study the optimal packet forwarding distance to minimize the hop count, and show that it is approximately 0.73R (for power attenuation degree 2). A hop count optimal, greedy, localized routing algorithm [referred as ideal hop count routing (IHCR)] for ad hoc wireless networks is then presented. We present another algorithm called expected progress routing with acknowledgment (referred as aEPR) for ad hoc wireless networks. Two variants of aEPR algorithm, namely, aEPR-1 and aEPR-u are also presented. Next, we propose projection progress scheme, and its two variants, 1-Projection and u-Projection. Iterative versions of aEPR and projection progress attempt to improve their performance. We then propose tR-greedy routing scheme, where packet is forwarded to neighbor closest to destination, among neighbors that are within distance tR. All described schemes are implemented, and their performances are evaluated and compared.     

Sparse structures for searching and broadcasting algorithms over internet graphs and peer-to-peer computing systems

In a broadcasting problem, a message is sent from a source to all the other nodes in the network. Blind flooding is a classical mechanism for broadcasting, where each node retransmits received message to all its neighbors. Despite its important advantages, an increase in the number of requests or the network size or density produces communication overheads that limit the scalability of blind flooding, especially in networks with dynamic topologies. Theoretically optimal solution is based on minimal spanning trees (MST), but its construction is expensive. We discuss here protocols based on local knowledge and newly proposed sparse structures. In weighted RNG (Relative Neighborhood Graph), messages are forwarded only on links which are not the ‘longest’ in any triangle. In weighted RNGQ, messages are forwarded to links which are not the longest in any triangle or quadrangle. Each node constructs weighted MST based on its 2-hop knowledge. Weighted LMST (Localized LMST) preserves only edges that are selected by both endpoints, and messages are forwarded on these links. Any available metric, such as delay, reliability, reputation etc. can be used as weight. Analysis and simulation show that weighted RNG performs better than existing Flooding and Rumor Mongering (or Gossip) schemes. The parameterless weighted LMST, MST, RNG and RNGQ (RNG over Quadrangle) based broadcasting schemes are compared, showing overall advantage for LMST. We also hint that a number of existing protocols can be simplified by applying concept from this article.     

Power-aware localized routing in wireless networks

A cost aware metric for wireless networks based on remaining battery power at nodes was proposed for shortest-cost routing algorithms, assuming constant transmission power. Power-aware metrics, where transmission power depends on distance between nodes and corresponding shortest power algorithms were also proposed. We define a power-cost metric based on the combination of both node's lifetime and distance-based power metrics. We investigate some properties of power adjusted transmissions and show that, if additional nodes can be placed at desired locations between two nodes at distance d, the transmission power can be made linear in d as opposed to d^α dependence for α ⩾ 2. This provides basis for power, cost, and power-cost localized routing algorithms where nodes make routing decisions solely on the basis, of location of their neighbors and destination. The power-aware routing algorithm attempts to minimize the total power needed to route a message between a source and a destination. The cost-aware routing algorithm is aimed at extending the battery's worst-case lifetime at each node. The combined power-cost localized routing algorithm attempts to minimize the total power needed and to avoid nodes with a short battery's remaining lifetime. We prove that the proposed localized power, cost, and power-cost efficient routing algorithms are loop-free and show their efficiency by experiments