Nowadays, Intelligent Transportation Systems (ITS) are known as powerful solutions for handling traffic-related issues. ITS are used in various applications such as traffic signal control, vehicle counting, and automatic license plate detection. In the special case, video cameras are applied in ITS which can provide useful information after processing their outputs, known as Video-based Intelligent Transportation Systems (V-ITS). Among various applications of V-ITS, automatic vehicle speed measurement is a fast-growing field due to its numerous benefits. In this regard, visual appearance-based methods are common types of video-based speed measurement approaches which suffer from a computationally intensive performance. These methods repeatedly search for special visual features of vehicles, like the license plate, in consecutive frames. In this paper, a parallelized version of an appearance-based speed measurement method is presented which is real-time and requires lower computational costs. To acquire this, data-level parallelism was applied on three computationally intensive modules of the method with low dependencies using NVidia’s CUDA platform. The parallelization process was performed by the distribution of the method’s constituent modules on multiple processing elements, which resulted in better throughputs and massively parallelism. Experimental results have shown that the CUDA-enabled implementation runs about 1.81 times faster than the main sequential approach to calculate each vehicle’s speed. In addition, the parallelized kernels of the mentioned modules provide 21.28, 408.71 and 188.87 speed-up in singularly execution. The reason for performing these experiments was to clarify the vital role of computational cost in developing video-based speed measurement systems for real-time applications.

Video-based vehicle speed measurement systems are known as effective applications for Intelligent Transportation Systems (ITS) due to their great development capabilities and low costs. These systems utilize camera outputs to apply video processing techniques and extract the desired information. This paper presents a new vehicle speed measurement approach based on motion detection. Contrary to feature-based methods that need visual features of the vehicles like license-plate or windshield, the proposed method is able to estimate vehicle’s speed by analyzing its motion parameters inside a pre-defined Region of Interest (ROI) with specified dimensions. This capability provides real-time computing and performs better than feature-based approaches. The proposed method consists of three primary modules including vehicle detection, tracking, and speed measurement. Each moving object is detected as it enters the ROI by the means of Mixture-of-Gaussian background subtraction method. Then by applying morphology transforms, the distinct parts of these objects turn into unified filled shapes and some defined filtration functions leave behind only the objects with the highest possibility of being a vehicle. Detected vehicles are then tracked using blob tracking algorithm and their displacement among sequential frames are calculated for final speed measurement module. The outputs of the system include the vehicle’s image, its corresponding speed, and detection time. Experimental results show that the proposed approach has an acceptable accuracy in comparison with current speed measurement systems.

In large data sets data pre-processing always has been the most essential data processing stages. Sampling and using small volumes of data has been an integrated part of data pre-processing to decrease training errors and increase speed of learning. In this study, instead of sampling from all data and using small parts of them, a method has been proposed to not only benefit from sampling but all data be used during training process. In this way, outliers would be detected and even used in completely different way. Using artificial neural networks, new features for instances will be built and the problem of intrusion detection will be mapped as a 10- feature problem. In fact, such a classification is for feature creation and as features in new problem only have discrete values, in final classification decision tree will be used. The results of proposed method on KDDCUP’99 datasets and Cambridge datasets show that this has improved classification in many classes dramatically.

In this paper, a hybrid classifier using fuzzy clustering and several neural networks has been proposed. With using the fuzzy C-means algorithm, training samples will be clustered and the inappropriate data will be detected and moved to another dataset (Removed-Dataset) and used differently in the classification phase. Also, in the proposed method using the membership degree of samples to the clusters, the class of samples will be changed to the fuzzy class. Thus, for example in KDD cup99 dataset, any sample will have 5 membership degrees to classes DoS, Probe, Normal, U2R, and R2L. Afterwards, the neural networks will be trained by new labels then using a combination of regression and classification methods, the hybrid classifier will be created. Also to classify the outlier data, a fuzzy ARTMAP neural network is employed which is a part of the hybrid classifier.
Evaluation of the proposed method is performed by KDDCup99 dataset for intrusion detection and Cambridge datasets for traffic classification problems. Our experimental results indicate that the proposed system has performed better than the previous works in the case of precision, recall and f-value also detection and false alarm rate. Also, ROC curve analysis shows that the proposed hybrid classifier has been better than the famous non-hybrid classifiers.