Kwangyong Lim

Real-time detection of speed-limit traffic signs on the real road using Haar-like features and boosted cascade

Along with the development of the intelligent vehicle, the Advanced Driver Assistance System(ADAS) has recently become an important issue. Traffic signs on the road provide crucial information to the driver. Recognizing all the traffic signs on the side of the road can be a difficult task for a driver who should watch the road ahead. To solve this problem, this paper proposes real-time detection methods using Haar-like features in a real road driving environment. We implement a reliable reduction method of the search area to improve the detection speed, masking methods and histogram equalization to improve the detection rate. The proposed method has shown higher detection rate and two times faster performance time than previous works.

Real-time traffic sign recognition based on a general purpose GPU and deep-learning

We present a General Purpose Graphics Processing Unit (GPGPU) based real-time traffic sign detection and recognition method that is robust against illumination changes. There have been many approaches to traffic sign recognition in various research fields; however, previous approaches faced several limitations when under low illumination or wide variance of light conditions. To overcome these drawbacks and improve processing speeds, we propose a method that 1) is robust against illumination changes, 2) uses GPGPU-based real-time traffic sign detection, and 3) performs region detecting and recognition using a hierarchical model. This method produces stable results in low illumination environments. Both detection and hierarchical recognition are performed in real-time, and the proposed method achieves 0.97 F1-score on our collective dataset, which uses the Vienna convention traffic rules (Germany and South Korea).

Vehicle License Plate Detection in Road Images

This paper proposes a vehicle license plate detection method in real road environments using 8 bit-MCT features and a landmark-based Adaboost method. The proposed method allows identification of the potential license plate region, and generates a saliency map that presents the license plates location probability based on the Adaboost classification score. The candidate regions whose scores are higher than the given threshold are chosen from the saliency map. Each candidate region is adjusted by the local image variance and verified by the SVM and the histograms of the 8bit-MCT features. The proposed method achieves a detection accuracy of 85% from various road images in Korea and Europe.

An Illumination Invariant Traffic Sign Recognition in the Driving Environment for Intelligence Vehicles

This paper proposes a traffic sign recognition method in real road environments. The video stream in driving environments has two different characteristics compared to a general object video stream. First, the number of traffic sign types is limited and their shapes are mostly simple. Second, the camera cannot take clear pictures in the road scenes since there are many illumination changes and weather conditions are continuously changing. In this paper, we improve a modified census transform(MCT) to extract features effectively from the road scenes that have many illumination changes. The extracted features are collected by histograms and are transformed by the dense descriptors into very high dimensional vectors. Then, the high dimensional descriptors are encoded into a low dimensional feature vector by Fisher-vector coding and Gaussian Mixture Model. The proposed method shows illumination invariant detection and recognition, and the performance is sufficient to detect and recognize traffic signs in real-time with high accuracy.

Real-time illumination-invariant speed-limit sign recognition based on a modified census transform and support vector machines

In this paper, we propose a robust illumination system for speed-limit sign recognition in real-time. Real-time traffic sign detection with various illuminations is one of the challenges in a vision-based intelligent vehicle system, as illumination varies greatly in real-world road images based on factors such as driving time, weather, lighting conditions, and driving directions. Our method uses a MCT (Modified Census Transform) as an illumination-invariant method for the real-time detection of traffic signs and uses a SVM (Support Vector Machine) as a classifier for detection and validation. With the proposed method, we have obtained a very high detection rate of 99.8% and recognition rates of 98.4% on various real-world driving images.

Real-time background subtraction based on GPGPU for high-resolution video surveillance

Demand for intelligent surveillance has been increasing, to automatically detect and prevent dangerous situations with surveillance cameras. Image analysis, the most essential element in intelligent surveillance system, has continuously developed and contributed to the improvement. To analyze surveillance videos, foreground segmentation is vital which require background modeling. This paper proposes background modeling method which is robust to illumination variation and shadow area. Also, the proposed method is applicable to high-resolution videos in real time with modification for GPU implementation. We validate our method on different types of dataset including our new benchmark dataset to analyze the result quantitatively and qualitatively. The execution time of proposed method is 228.2 FPS for High Definition videos with NVIDIA GTX660.

Illumination invariant color segmentation method based on cluster center tree for traffic sign detection

This paper proposes a color segmentation method that can locate candidate regions of traffic signs accurately and reliably from real world images. In the real world, there are various light conditions which make the color segmentation very difficult problem. Hence, we propose an illumination invariant color segmentation method. The proposed method consists of two parts; 1) cluster center tree-based segmentation 2) illumination estimation. Cluster center tree is trained for color segmentation. Illumination estimation algorithm classifies light condition of the input images. We validate the proposed method qualitatively and quantitatively with 1,745 images containing red and blue traffic signs captured with four light conditions; sunny, cloudy, rainy and night. The proposed method achieves the high detection rate of 99.25% in sunny, 98.33% in cloudy, 87.85% in rainy and 88.70% at night.

Posture Recognition for a Bi-directional Participatory TV Program based on Face Color Region and Motion Map

As intuitive hardware interfaces continue to be developed, it has become more important to recognize the posture of the user. An efficient alternative to adding expensive sensors is to implement computer vision systems. This paper proposes a method to recognize a users postured in a live broadcast bi-directional participatory TV program. The proposed method first estimates the position of the users hands by generation a facial color map for the user and a motion map. The posture is then recognized by computing the relative position of the face and the hands. This method exhibited 90% accuracy in an experiment to recognize three defined postures during the live broadcast bi-directional participatory TV program, even when the input images contained a complex background.

Real-time traffic sign detection with vehicle camera images

This paper presents a real-time traffic sign detection method using color properties and shape-based features for real-world environment applications. The proposed method has two main steps: color-based region segmentation and shape-based verification of the segmented area. In the first step, region-of-interest (ROI) is roughly segmented by a simple color-based thresholding method and each segment is corrected by a guided filter. Next, each ROI is verified through a shape analysis to decide whether the ROI is a circle or a triangle. For detecting circles, we compare three different methods: RSD, BCT, and STVUE. For triangles, RPD, VBT and STVUT were applied. We evaluated these alternatives with 232 experimental images containing 142 circular signs and 82 triangular signs. We found that RSD in the circle detection and STVUT in the triangle detection provide the best detection rates of 93% and 90% respectively. The main contribution of this paper is to present a novel approach for extracting boundary of traffic sign.

Real-time Korean traffic sign detection and recognition

In this paper, we propose a real-time Korean traffic sign detection and recognition method based on color properties and shape geometries of images. The proposed method supports detecting and recognizing various shapes of traffic signs in real-time. Our method consists of four stages: 1) color based image segmentation; 2) region of interest (ROI) detection; 3) shape classification; and 4) numeral recognition. The proposed method can classify even the signs that are partially occluded. In addition, we improve efficiency of shape classification by using simple shape geometry measurements. Our experiment shows that our approach can provide high classification accuracies for octagonal shape signs (92%) and speed-limit signs (94.5%).