Current Automatic License Plate Recognition (ALPR) systems are designed to work on license plates (LP) from specific countries and use country-specific information which limits their practical applicability. This study presents a deep ALPR system designed to be applicable to multinational LPs. The proposed approach consists of three main steps – LP detection, unified character recognition, and multinational LP layout detection. The system is mainly based on the you only look once (YOLO) networks. Particularly, tiny YOLOv3 was used for the first step whereas the second step uses YOLOv3-SPP. The localized LP is fed into YOLOv3-SPP for character recognition. A layout detection algorithm is proposed that can extract the correct sequence of LP numbers from multinational LPs. The proposed system was evaluated on LP datasets from five countries and consumes about 42 ms per image on average for extracting LP number.
2. Automatic trimap generation and artifact reduction
Image Matting is an actively researched topic and refers to the accurate extraction of the foreground from an image. It is a severely under-constrained problem, hence, a user input in the form of a trimap is required. However, users must generate this trimap manually, which is an exhausting and time-consuming process. To tackle this issue, we propose a simple yet effective approach to generate optimal trimaps automatically by combining image saliency, graph cut segmentation (lazy snapping), and fuzzy c-means clustering (FCM). Lazy snapping is an interactive segmentation technique that requires foreground and background scribbles as input. Instead of using user-assisted foreground scribbles, we utilize a saliency map as foreground scribbles and input it to the lazy snapping. This results in a coarsely segmented foreground object. To generate an optimal trimap, we locally cluster the boundary region of the foreground segmentation using FCM. We tested our algorithm on alpha matting evaluation and salient object datasets. In addition to generating accurate trimaps automatically, the alpha mattes generated by our optimal trimaps contain fewer artifacts and were computed faster as compared to previous works. Finally, our approach does not rely on depth data like the previous methods.
3. Automatic Detection System of Deteriorated PV Modules Using Drone with Thermal Camera
The output efficiency of photovoltaic (PV) power stations deteriorates with the passage of time due to factors like hotspots, shaded cell or module, short-circuited bypass diodes, etc. Researchers have proposed the use of drones equipped with thermal cameras for PV power station monitoring. However, most of these drone-based approaches require technicians to manually control the drone which could be demanding. To tackle this issue, this study presents an autonomous drone-based solution which can automatically detect and estimate the exact location of faulty PV modules in the power station. In addition, an automatic drone flight path planning algorithm is proposed which eliminates the requirement of manual drone control. The system also utilizes an image processing algorithm to process RGB and thermal images for fault detection. The system was evaluated on a 1-MW solar power plant located in Suncheon, South Korea.
4. Allergic Rhinitis Prediction based on Artificial Intelligence
A system based on Artificial Intelligence (AI) was developed to predict the likelihood of allergic rhinitis. The system can learn from weather and user data for a personalized prediction. Artificial Neural Network was used to predict the likelihood for patients who get affected immediately while Recurrent Neural Network was used for patients who get affected several days after bad weather conditions.
1. Vision-based Computer Mouse Control
This project can control a computer's mouse via user's pupil movement and eye blinking/winking. The target audience for this system are physically challenged people who are unable to use computer in the conventional way (mouse/keyboard). Users can move the mouse cursor via their pupil movements and can also clock left and right mouse button via eye winks.
2. Bike Fit Calculator
This project aims at improving the performance and efficiency of cyclists and preventing injuries. Bike fitting is the process of adjusting a bike for a cyclist in order to achieve optimal performance and efficiency. Incorrectly setting up a cycle can lead to injuries such as cyclist's palsy, cyclist back, and anterior knee pain. The developed system calculates the optimal saddle height for cyclists to avoid knee injuries. The system tracks markers on the cyclists and calculates the optimal saddle height by using various angle measurements.
3. Marker Tracking-based Gait Analysis
This project provides useful information which is helpful in gait analysis. The system provides hip, knee, and ankle angles as the person walks in a video. The person in the video must wear markers on his joints which will be tracked in the video. Due to the occlusion problem, the location of the hip is estimated using the knee marker and a marker placed slightly above the knee. A GUI was also developed for making the system easy to use for a layman. The project output a .csv file with various data and a video file embed with graphs for the joint angles.