Browsing by Author "Tsai, Chia-Chi"

Now showing 1 - 13 of 13

A Blind Spot Detection Warning System based on Gabor Filtering and Optical Flow for E-mirror Applications
(2018-01-01)
Chang, Shun-Min
;
Tsai, Chia-Chi
;
Guo, Jiun-In
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
Blind Spot Detection (BSD) is an important technique for ADAS. We propose a BSD algorithm using Gabor filtering and optical flow to detect vehicles in the blind spot region for both day-time and night-time applications. For the day-time scene, the Gabor filtering is used to detect the vehicles, inside lane line, and outside lane line. After detection, the optical flow information calculated according to Horn-Schunck method is used to judge the motion of the vehicle candidates and filter the mistake-judgement. For the night-time scene, we try to find the head-light of the approaching cars. First, we perform binarization on the image first, find the center of gravity of the light-area, classify the light-area into 2 groups and judge it as a vehicle or not. The proposed BSD system achieves 93.58% recall and 95.83% precision in day time scene and 90.22% recall and 92.76% precision in night time scene. The algorithm can achieve performance of 89 fps on Intel Core I7 and 50 fps on Renesas R-Car M2 under 640x480 resolution.
A Multiple-Lane Vehicle Tracking Method for Forward Collision Warning System Applications
(2017-01-01)
Li, Yuan-Fu
;
Tsai, Chia-Chi
;
Lai, Yi-Ting
;
Guo, Jiun-In
;
電子工程學系及電子研究所
This paper proposes a vehicle tracking method for the Forward Collision Warning System (FCWS). The proposed method applies vehicle detection and feature tracking to each frame in the videos to enhance the detection accuracy and increase the stability of vehicle localization. The proposed method can be applied to three-lane FCWS which includes left, main, and right lanes. The performance of the proposed three-lane FCWS can achieve 720x480 video@183 fps in average when realized in PC with Intel Core i7 processor with average detection accuracy of 94.05% at daytime and 86.90% at night. The proposed method is also implemented on Freescale i.MX6 embedded platform with a USB webcam to capture the video. Under the D1 (720x480) resolution, the performance of the proposed 3-lane FCWS can achieve 29 fps.
A Vision Radar System for Car Safety Driving Applications
(2017-01-01)
Tsai, Chia-Chi
;
Lai, Yi-Ting
;
Li, Yuan-Fu
;
Guo, Jiun-In
;
電子工程學系及電子研究所
This paper presents a vision radar system (VRS) integrating Lane Departure Warning System (LDWS) and Forward Collision Warning System (FCWS) for car safety driving applications. A dynamic threshold, which improves the accuracy of lane and vehicle detection in various weather conditions, and multiple frame approval, which conquers the effect of some tough scenes, are proposed. The designed radar-map facilitates warning message issuing of safety distance for forward and potential cutin vehicles. The proposed VRS can achieve 22 fps at 720x480 (D1) resolution on Freescale i.MX6 embedded platform.
Dark channel prior based video dehazing algorithm with sky preservation and its embedded system realization for ADAS applications
(2019-04-29)
Tsai, Chia-Chi
;
Lin, Cheng-Yen
;
Guo, Jiun-In
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
Dark Channel Prior (DCP) is one of the significant dehazing methods based upon the observation of the key features of the haze-free images. But it has disadvantages; high computational complexity, over-enhancement in the sky region, flickering artefacts in video processing, and poor dehazing. Therefore, we propose improved solutions to solve the aforementioned drawbacks. First, we adopt the fast one-dimensional filter, look-up table, and program optimization to reduce the computational complexity. Next, we follow by using a part of the guided filter for sky detection and to preserve the sky region from noise by avoiding over recovery. Then, we propose an airtight update strategy and adjust the radius of a guided filter to reduce the flickering artifacts and also propose an airtight estimation method to produce the better dehazing result as the final step of our algorithm. The improved results from our proposed algorithm are stable and are obtained from the real-time processing suitable for ADAS, surveillance, and monitoring systems. The implementation of the proposed algorithm has yielded a processing speed of 75 fps and 23 fps respectively on an NVIDIA Jetson TX1 embedded platform and Renesas R-Car M2. both on DI (720x480) resolution videos. (C) 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
Design and Realization of Delta-Sigma Analog-to-Digital Converter in LTPS Technology
(2009)
Tsai, Chia-Chi
;
Ker, Ming-Dou
;
Li, Yu-Hsuan
;
Kuo, Chung-Hung
;
Li, Chun-Huai
A delta-sigma analog-to-digital (A/D) converter designed and implemented with the thin-film transistors (TFTs) on glass substrate is proposed, which has been successfully verified in a 3-mu m low-temperature polycrystalline silicon (LTPS) process. The experimental results have shown that the probability in the bit stream of the digital output from the delta-sigma modulator is correctly fit in the analog input voltage ratio. Such a bit stream can be converted into 8-bit digital code successfully under the operation voltage of 10V with thin film transistors (TFTs) on the glass substrate. The proposed delta-sigma AID converter can be further used to achieve the precise analog circuits for System-on-Panel (SOP) or System-on-Glass (SOG) applications, which enables the analog circuits to be integrated in the active matrix LCD (AMLCD) panels.
Embedded Multiple Object Detection Based on Deep Learning Technique for Advanced Driver Assistance System
(2017-01-01)
Chang, Fong-An
;
Tsai, Chia-Chi
;
Tseng, Ching-Kan
;
Guo, Jiun-In
;
電子工程學系及電子研究所
This paper proposes an optimized pedestrian and vehicle detection method based on deep learning technique. We optimize the convolutional neural network architecture by three mainly methods. The first one is the choice of the learning policy. The second one is to simplify the convolutional neural network architecture. The last one is careful choice of training samples. With limited loss of accuracy, we can greatly speed up the original deep learning method coming from CAFFE. The proposed system is developed on PCs and implemented on the platforms of both the PC and embedded systems. We can achieve around 90% accuracy when it is tested on an open-source dataset. On PCs with Intel i7@3.5GHz CPU, the proposed design can reach the performance about 720x480 video at 25 frames per second. On the NVIDIA JETSON TX1 embedded system, the proposed design can reach the performance about 720x480 video at 5 frames per second.
Implementation of delta-sigma analog-to-digital converter in LTPS process
(2010-11-01)
Tsai, Chia-Chi
;
Wang, Tzu-Ming
;
Ker, Ming-Dou
;
電機學院
;
College of Electrical and Computer Engineering
An on-panel delta-sigma analog-to-digital converter (ADC) has been implemented and verified for 3-mu m low-temperature polysilicon (LTPS) technology with two basic blocks: a delta-sigma modulator and a decimation filter. From the experimental results, the digital output from the delta-sigma modulator is correctly matched with the analog input voltage ratio such that the digital output can be converted into 8-bit digital code successfully under a supply voltage of 10 V from the decimation filter. The implemented on-panel delta-sigma ADC can be used for the application of temperature-to-digital converter on glass substrate.
Implementation of delta-sigma analog-to-digital converter in LTPS process
(2010-11-01)
Tsai, Chia-Chi
;
Wang, Tzu-Ming
;
Ker, Ming-Dou
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
An on-panel delta-sigma analog-to-digital converter (ADC) has been implemented and verified for 3-mu m low-temperature polysilicon (LTPS) technology with two basic blocks: a delta-sigma modulator and a decimation filter. From the experimental results, the digital output from the delta-sigma modulator is correctly matched with the analog input voltage ratio such that the digital output can be converted into 8-bit digital code successfully under a supply voltage of 10 V from the decimation filter. The implemented on-panel delta-sigma ADC can be used for the application of temperature-to-digital converter on glass substrate.
LiDAR/Camera Sensor Fusion Technology for Pedestrian Detection
(2017-01-01)
Wu, Tai-En
;
Tsai, Chia-Chi
;
Guo, Jiun-In
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
Nowadays, the machine learning for object detection is growing popular and widely adopted in many fields, such as surveillance, automotive, passenger flow analysis, etc. This paper focuses on the research of Lidar/camera sensor fusion technology for pedestrian detection to ensure extremely high detection accuracy. In order to reduce the false-positive rate and the object occlusion problem, which usually happened in camera-based pedestrian detection, we use 3D point cloud returning from Lidar depth sensor to do the further examination on the object's shape. The proposed Lidar/camera sensor fusion design complements the advantage and disadvantage of two sensors such that it is more stable in detection than others. The region proposal is given from both sensors, and candidate front two sensors are also going to the second classification for double checking. It maximums the detection rate and achieves average 99.16% detection accuracy for pedestrian detection.
One Stage Detection Network with an Auxiliary Classifier for Real-Time Road Marks Detection
(2018-01-01)
Lin, Guan-Ting
;
Santoso, Patrisia Sherryl
;
Lin, Che-Tsung
;
Tsai, Chia-Chi
;
Guo, Jiun-In
We construct a robust road mark detector that achieves high accuracy with real-time processing performance ( 32 fps) under nVidia Titan-X GPU. We combine one stage deep learning detector with auxiliary CNN classifiers as a robust road marks detector. We found out that one stage detector not only detects multiple objects via single inference efficiently, but also remains a good accuracy in performance perspective. However, to make it better, we add an extra CNN classifier as the back part of the proposed architecture to reduce false positive and get better accuracy. The proposed detector can achieve 86.8% mAP in our in-house six-class road mark database.
Stop Line Detection and Distance Measurement for Road Intersection based on. Deep Learning Neural Network
(2017-01-01)
Lin, Guan-Ting
;
Santoso, Patrisia Sherryl
;
Lin, Che-Tsung
;
Tsai, Chia-Chi
;
Guo, Jiun-In
In this paper, we introduce Boost-CNN, a robust stop-line detector that can detect objects (stop line) with competitive tradeoff between speed and accuracy. Boost-CNN consists of an AdaBoost classifier and a CNN. The former is our region proposal generator and it is further combined with the later to be a stop-line detector. In addition, an automatic hard mining method is proposed to reduce the number of false alarm. Our proposed detector achieves 91.5% in accuracy and has 100 FPS performance in test time (performed on NVITAA DIGITS DevBox and Titan X GPU).
Summary Embedded Deep Learning Object Detection Model Competition
(2019-01-01)
Guo, Jiun-In
;
Tsai, Chia-Chi
;
Yang, Yong-Hsiang
;
Lin, Hung-Wei
;
Wu, Bo-Xun
The embedded deep learning object detection model competition in IEEE MMSP2019 focuses on the object detection for sensing technology in autonomous driving vehicles, which aims at detecting small objects in worse conditions through embedded systems. We provide a dataset with 89,002 annotated images for training and 1,500 annotated images for validation. We test participants' models through 6,000 testing images, which are separated into 3,000 for qualification and 3,000 for finals. There are 87 teams of participants registered this competition and 14 teams submitted the team composition. At last there are nine teams entering the final competition and five teams submitting their final models that can be realized in NVIDIA Jetson TX-2. At the end, only one team's model passed the target accuracy requirement for grading and became the champion of the contest, which the winner is team R.JD.
Vehicle Detection and Classification based on Deep Neural Network for Intelligent Transportation Applications
(2018-01-01)
Tsai, Chia-Chi
;
Tseng, Ching-Kan
;
Tang, Ho-Chia
;
Guo, Jiun-In
;
電子工程學系及電子研究所
This paper proposes an optimized vehicle detection and classification method based on deep learning technology for intelligent transportation applications. We optimize the Convolutional Neural Network (CNN) architecture by fine-tuning the existing CNN architecture for the intelligent transportation applications. The proposed design achieves the accuracy of miss rate around 10% when FPPI is 0.1. Realized on nVidia Titan-X GPU, the proposed design can reach the performance about 720x480 video under different weather condition (day, night, raining) at 25fps. The proposed model can achieve 90% accuracy on three target vehicle classes including small vehicles (Sedan, SUV, Van), big vehicles (Bus) and Trucks.