This paper presents a novel traffic sign recognition system comprising of: (i) Color/shape classification, (ii) Pictogram extraction, (iii) Features selection and, (iv) Lyapunov Theory-based Radial Basis Function neural network (RBFNN). In the proposed system, traffic signs are first segmented and classified with regard to its unique color and shape in order to partition a large set of data into smaller subclasses. Within these subclasses, all redundant information except the pictogram is discarded for feature selection since the pictogram contains critical information for road users. Principle Component Analysis (PCA) is applied to extract salient points for traffic sign dimensionality reduction. This is followed by the Fisher's Linear Discriminant (FLD) to further obtain the most discriminant features. These features are fed into RBFNN for training with a proposed weight updating scheme based on Lyapunov stability theory. The performance of the proposed system is evaluated with Malaysian road signs with promising recognition rate.

Intra color-shape classification for traffic sign recognition

In this paper we propose a traffic sign recognition system using an on-board single camera for Advanced Driver Assistance Systems (ADAS), including detection, recognition and tracking. We combine RGB ratios based color segmentation with automatic white balance preprocessing and Douglas-Peucker shape detection to establish ROIs. Scale and rotation invariant BRISK features are applied for recognition, matching the features of the candidates to those of template images that exist in database. Tracking-Learning-Detection (TLD) framework is adopted to track the recognized signs in real time to provide enough information for driver assistance function. This paper presents lots of experiments in real driving conditions and the results demonstrate that our system can achieve a high detection and recognition rate, and handle large scale changes, motion blur, perspective distortion and various illumination conditions as well.

Robust traffic sign recognition and tracking for Advanced Driver Assistance Systems

A hybrid traffic sign recognition scheme combining of knowledge-based analysis and radial basis function neural classifier (RBFNN) is proposed in this paper. Initially, traffic signs are detected from the road scenes using color segmentation method. The extracted signs are then passed to the recognition system for classification. The proposed recognition technique composes of three stages: (i) color histogram classification, (ii) shape classification and, (iii) RBF neural classification. Based on the unique color and shape of traffic signs, they can be classified into smaller subclasses and can be easily recognized using RBFNN. Before feeding traffic sign into the RBFNN, traffic sign features are extracted by Principle Component Analysis (PCA) in order to reduce the dimensionality of the original images. This is followed by the Fisher's Linear Discriminant (FLD) to further obtain the most discriminant features. The performance of the proposed hybrid system is evaluated and compared to the purely neural classifier. The experimental results demonstrate that the proposed method has better recognition rate.

New hybrid technique for traffic sign recognition

Spatial operations such as spatial join combine two objects on spatial predicates. It is different from relational join because objects have multi dimensions and spatial join consumes large execution time. Recently, many researches tried to find methods to improve the execution time. Parallel spatial join is one method to improve the execution time. Comparison between objects can be done in parallel. Spatial datasets are large. R-Tree data structure can improve the performance of spatial join. In this paper, a parallel spatial join on Graphic processor unit (GPU) is introduced. The capacity of GPU which has many processors to accelerate the computation is exploited. The experiment is carried out to compare the spatial join between a sequential implementation with C language on CPU and a parallel implementation with CUDA C language on GPU. The result shows that the spatial join on GPU is faster than on a conventional processor.

/pdf/spatial-join-with-r-tree-on-graphics-processing-units-34w8serz2e.pdf

Spatial Join with R-Tree on Graphics Processing Units

SUMMARY

Efficient spatial index is essential for querying spatial sensor nodes in the context of smart city. Sensor nodes are usually unevenly distributed in real situations. In this setting, R-tree and its variants may cause large overlap and coverage among branch nodes, which impact the query efficiency greatly. To address this challenge, this paper proposes a novel skewness-aware clustering tree (SWC-tree) by clustering sensor nodes. Sensor nodes in a dense region will be put into the same node. Thus, overlap and coverage among node regions are less than that of R-tree and its variants. As dense regions contain more sensor nodes, we assign a higher priority to these region nodes for facilitating the query operation. Experimental results show that in the context of skewed distribution, SWC-tree is efficient in performance for conducting insertion, deletion, and query operations of sensor nodes. Copyright © 2012 John Wiley & Sons, Ltd.

Skewness‐aware clustering tree for unevenly distributed spatial sensor nodes in smart city

In order to solve these problems how to easily generate candidate frequent item sets and fast compute support of candidate item sets, an algorithm of association rules mining based on binary has been introduced. However, one presented binary mining algorithm is only suitable for mining some relative short frequent item sets since the way of generating candidate item sets is also similar to apriori, another is only suitable for mining long frequent item sets, which generates candidate item sets by up-down search strategy. And so aiming to mining general frequent item sets, this paper proposes an algorithm of association rules double search mining based on binary, which is different from tradition association rules mining algorithm based on double search strategy. The algorithm doesnpsilat use combination of set theory to generate candidate item sets but binary logic operation that is also used to compute support of candidate item sets, which can use character digital to reduce the number of scanned transaction. The algorithm gets rid of shortage about some presented algorithms based on binary. The experiment based on above three algorithms indicates that the efficiency about double search strategy is fast and efficient when mining general frequent item sets which arenpsilat confined.

An algorithm of association rules double search mining based on binary

The balance of data and the utilization of resources are essential to distributed spatial database system. The paper presents an efficient parallel spatial query algorithm which takes seriously the organization of spatial data into account. The algorithm adopts a balanced spatial data partitioning strategy for distributed spatial databases. According to the characteristics of data partitioning, it builds a packing R-tree as its index. The strategy also considers the problem of computing distribution. By replicating index to every site, each site can access different entry in the same index node at the same time. Based on the organization of spatial data, the algorithm can simultaneously execute query operation at different site in both filtration phase and refinery phase. So it obviously improves spatial query performances. In order to solve multiple paths search problem caused by R-tree index, the algorithm brings in globe stack to buffer temporary index nodes. It settles the difficult problem flexibly in distributed spatial databases. For simplicity, the paper discusses the parallel algorithm in 2-dimensional space. Through the experiments conducting on many real datasets, it shows better performance in various spatial query operations.

A new parallel spatial query algorithm for distributed spatial databases

This paper proposes an algorithm of objective association rules used in data mining, which is suitable for mining some potential rules related to the given object from large data warehouse. The basic principle of algorithm is that the association rules were formed according to this especial information, which is mined with binary logic "and" operation from many attribute-values of database, where data have been expressed by binary. The theory of algorithm is quite simple. It is feasible and efficient to execute the algorithm. It is a practical significance to apply to the algorithm in some applicable domains to search especial objective association rules.

The Algorithm of Objective Association Rules Mining Based on Binary

In this paper, in order to reduce the times of scanning database when presented algorithms compute support of candidate frequent itemsets, in order to improve the method of computing support of candidate frequent itemsets, and in order to further improve the efficiency of algorithm, based on up search strategy of Apriori, we propose an algorithm of association rules mining based on sequence number. The algorithm would use the method of binary Boolean calculation to generate candidate frequent itemsets of binary form, and gain support of candidate frequent itemsets by computing Sequence Number Degree (SND), which is gained through computing these Sequence Number (SN) of all these items contained by candidate frequent itemsets. The algorithm only need scan once all these transactions in database to indeed improve the efficiency of algorithm. The experiment indicates the efficiency of this algorithm is faster and more efficient than presented algorithms.

An algorithm of improved association rules mining

A theoretical study for modeling technique of the remote sensing image classification based on the minimum description length (MDL) principle is presented in the paper. According to the MDL principle, modeling problem is an optimization procedure to find the shortest expected code length. Kullback-Leibler (KL) divergence is adopted as the system cost function to measure expected codelength, and the codelength will be the model we desired. The advantage of using the MDL principle to build appropriate model is analyzed theoretically, model optimization technique also is described.

Zu-Kuan Wei

Papers

An algorithm of association rules double search mining based on binary

A new parallel spatial query algorithm for distributed spatial databases

The Algorithm of Objective Association Rules Mining Based on Binary

An algorithm of improved association rules mining

Modeling multisource remote sensing image classifier based on the MDL principle: Theoretical aspects