Edwin Sybingco

Automated traffic violation apprehension system using genetic algorithm and artificial neural network

Developing countries face the problem of crowded and congested roads because of inefficient implementation of traffic rules. Motorists ignore the rules because they are not apprehended and can get away easily. This paper proposes an intelligent traffic system that is able to automatically detect and apprehend traffic violators, specifically motorists who either swerve or block the pedestrian lane. The system is designed by integrating three processes: violation detection, plate localization and plate recognition. The violation detection and plate localization were realized using genetic algorithm while the plate recognition process was performed using an artificial neural network. The recognition of the plate number is highly dependent on the position of the detected vehicle with respect to the camera. Thus, the recognized plate number will only be supplementary information about the violator; the physical attributes of the vehicle which is captured by the violation detection process will be the main information on the violator. Based on the results of 48 images tested, the overall system was able to detect the mentioned violations and to identify the plate number of the vehicles that were detected as traffic violators, with an average accuracy of 90.67%, and program runtime of 1.34 seconds.

Robust stock trading using fuzzy decision trees

Stock market analysis has traditionally been proven to be difficult due to the large amount of noise present in the data. Different approaches have been proposed to predict stock prices including the use of computational intelligence and data mining techniques. Many of these methods operate on closing stock prices or on known technical indicators. Limited studies have shown that Japanese candlestick analysis serve as rich information sources for the market. In this paper decision trees based on the ID3 algorithm are used to derive short-term trading decisions from candlesticks. To handle the large amount of uncertainty in the data, both inputs and output classifications are fuzzified using well-defined membership functions. Testing results of the derived decision trees show significant gains compared to ideal mid and long-term trading simulations both in frictionless and realistic markets.

Fuzzy logic based vehicular plate character recognition system using image segmentation and scale-invariant feature transform

This paper proposes a vehicle plate optical character recognition method using scale invariant feature transform integrated with image segmentation and fuzzy logic. Image segmentation separates every character in a plate area to get the features of every character obtained. Scale Invariant Feature Transform or SIFT on the other hand, allows the extraction of every feature of each character obtained from the plate. Fuzzy logic analyzes the features obtained from the SIFT algorithm which is proposed to detect the characters correctly. This program used MATLAB to determine the performance of the algorithm. Using the proposed algorithm, it was shown how the algorithm was effective on extracting plate character features as well as recognizing the characters in a given image. Results show that the algorithm has an accuracy of 90.75% and now ready to use for other implementation. This can be incorporated to present optical character recognition system and test its validity and accuracy for practical purposes.

Philippine vehicle plate localization using image thresholding and genetic algorithm

This paper proposes a vehicle plate localization method using genetic algorithm integrated with image thresholding. Image thresholding outputs a value which varies on the time the image is captured. Genetic algorithm on the other hand, executed the license plate region detection of the digital image which depends on the set-level of the image threshold values obtained. Using the proposed algorithm, it was shown how the algorithm was effective on finding the plate location in a given image. Results show that the different parameters tested were successful and converges to a point where the plate locations can be located. The algorithms were tested on an image of a vehicle equipped with a license plate on its frontal view tested on a large number of trials. The genetic algorithm initialized 2000 chromosomes as its initial population and a fixed generations count of 100. It was observed that the time it took for the program to locate the plate is about 3 seconds. Another finding observed is that by varying the initial chromosome count and generation count will lead to longer computation time with increased accuracy. On the contrary, if the initial values were lessened, computation time will be less but the accuracy lessen. Results show that this plate localization technique successfully locates the plate and may be calibrated depending on the time of analysis.

Vision based pedestrian detection using Histogram of Oriented Gradients, Adaboost & Linear Support Vector Machines

Pedestrian detection systems are valuable in a variety of applications such as in advanced driver assistance systems and advanced robots. This study presents a pedestrian detection system that uses Histogram of Oriented Gradients (HOG) as feature descriptor, and AdaBoost and Linear Support Vector Machines (SVM) as classifiers. The entire system is tested and evaluated in both publicly available databases and personally acquired videos. The pedestrian detection system has been tested and results show that it can detect pedestrians. Experiments showed that the system is up 20% faster compared to OpenCVs default detector.

Identification of river hydromorphological features using Viola-Jones Algorithm

In this paper, a quadcopter equipped with a camera was used to capture images from a river. These captured images were used as training data in the detection program used to detect the hydromorphological features in the area of the river such as trees, roofs, roads and the shore. The Viola-Jones Algorithm was used in order to detect, identify and recognize hydromorphological features due to its speed and simplicity of implementation. Testing was done using different images to verify the effectiveness of detection. System evaluation and success of the appropriateness of the Viola-Jones Algorithm was determined using the percentage of correct detected features in the image. The study showed that the Viola-Jones has shown that it is effective in detecting some features due to the complexity of the hydromorphological images.

A remote-controlled land and water traversing vehicle for flood search operation

A remote controlled robot that can traverse on land and water with human detection for flood search operation has been made to aid rescuers and volunteers in conducting search operations. The robot can traverse on water and different types of terrain, namely wooden floor, cement, rocks and mud. It can handle an additional load of less than 8 lbs and a water current of 0.385 meters per second. It can be controlled of up to a distance of 290 meters, line-of-sight. The human detection algorithm achieved an accuracy of 93.6 percent in terms of detecting unique ROI for every set of 15 frames.

GA Optimization of Coconut Sugar Cooking Process: A Preliminary Study using Stochastic Universal Sampling (SUS) Technique

This study presents the optimization of cooking process for the coconut sugar. Designing the process control of cooking coconut sugar requires dynamic programming that uses nonlinear differential equations which could be difficult to model and analyze. The development of the optimization process will make use of genetic algorithm (GA) based approach using stochastic universal sampling as its selection process. The developed system will be incorporated for the automation of coconut sugar production. Factors associated in the drying of coconut sap like treatment time and color of the honey were considered in this study.

A Neuro-Fuzzy Mixing Control Model for the Cooking Process of Coconut Sugar

This study presents a neuro-fuzzy system used in developing an appropriate model for the mixing control of the coconut sugar cooking process. The developed model is trained and tested using actual data and process gathered from cooking coconut sap run in several trials. Adaptive neuro-fuzzy inference system (ANFIS) was the primary tool used to model the control cooking process. Grid partition, subtractive clustering and fuzzy c-means clustering were used in the fuzzification of the training data. Then, the neural network generates the fuzzy rules for the model, which are evaluated to measure the performance of the model. Moreover, experimental results show the detailed comparison of the performance of each fuzzy model. Among the 3 training models used, the fuzzy c-means clustering provided the best performance with only 3 fuzzy rules extracted with an accuracy of 95.4% during testing.

Passenger demand forecast using optical flow passenger counting system for bus dispatch scheduling

This paper presents the development of a computer vision system for on-board bus passenger counting using optical flow. The system provides the demand forecast for the bus dispatch scheduling system in the EDSA route of Metro Manila. The camera system installed in the bus captures the images of boarding and alighting passengers. Optical flow is used for the motion detection and tracking of passengers. The passenger count for every bus are relayed to a central dispatch scheduling system, wherein a fuzzy logic controller determines the bus dispatch timetable. The results of the experiments conducted for bus passenger counting are presented in here.