Search Results

Facial expression, a significant way of nonverbal communication, effectively conveys humans' mental state, emotions and intentions. Understanding of emotions through these expressions is an easy task for human beings. However, when it comes to Human Computer Interface (HCI), it is a developing research field that enables humans' to interact with computers through touch, voice, and gestures. Communication through expression in HCI is still a challenge. In addition, there are a variety of fields such as automotive, biometric, surveillance, teleconferencing etc. in which expression recognition system can be applied. In recent years, several different approaches have been proposed fr facial expression recognition, but most of them work only under definite environmental conditions. The proposed framework aims to recognize expressions (by analyzing the facial features extracted) based on the Active Shape Model (ASM).

Dense depth estimation is a critical application in the field of robotics and machine vision where the depth perception is essential. Unlike traditional approaches which use expensive sensors such as LiDAR (Light Detection and Ranging) devices or stereo camera setup, the proposed approach for depth estimation uses a single camera mounted on a rotating platform. This proposed setup is an effective replacement to usage of multiple cameras, which provide around view information required for some operations in the domain of autonomous vehicles and robots. Dense depth estimation of local scene is performed using the proposed setup. This is a novel, however challenging task because baseline distance between camera positions inversely affect common regions between images. The proposed work involves dense two view reconstruction and depth map merging to obtain a reliable large dense depth map.

Autonomous braking systems are prevalent in mid/upper-mid range vehicles today. The major drawback: acute boundary condition during which the system will function. The paper describes the implementation of Adaptive Range LIDAR Systems (ARLS) containing a state of the art collimator and wave shaper with a 140̊ sweep MEMS mirror, capable of calculating beam convergence as a function of distance, considering multiple obstacles ahead of it. The paper also describes the use of ARLS for ACC (Adaptive Cruise Control) and Autonomous braking, reinforcing the available software structure with more data points. Contrary to the other systems that detect objects/obstacles from a stationary point of reference, ARLS determines the velocity of obstacle with respect to the ground point of reference and computes most optimum brake effort curve.

This paper presents a vision based pedestrian detection system. The presented algorithm is a novel method that accurately segments the pedestrian regions in real time. The fact that the pedestrians are always vertically aligned is taken into consideration. As a result, the edge image is scanned from bottom to top and left to right. Both the color and edge data is combined in order to form the segments. The segmentation is highly dependent on the edge map. Even a single pixel dis-connectivity would lead to incorrect segments. To improve this, a novel edge linking method is performed prior to segmentation. The segmentation would consist of foreground and background segments as well. The background clutter is removed based on certain predefined conditions governed by the camera features. A novel edge based head detection method is proposed for increasing the probability of pedestrian detection. The combination of head and leg pattern will determine the presence of pedestrians.

Advanced Driver Assistance Systems (ADAS) have become an inevitable part of most of the modern cars. Their use is mandated by regulations in some cases; and in other cases where vehicle owners have become more safety conscious. Vision / camera based ADAS systems are widely in use today. However, it is to be noted that the performance of these systems is depends on the quality of the image/video captured by the camera. Low illumination is one of the most important factors which degrades image quality. In order to improve the system performance under low illumination, it is required to first enhance the input images/frames. In this paper, we propose an image enhancement algorithm that would automatically enhance images to a near ideal condition. This is accomplished by mapping features taken from images acquired under ideal illumination conditions on to the target low illumination images/frames.

Camera sensors that are made of silicon photodiodes and used in ordinary digital cameras are sensitive to visible as well as Near-Infrared (NIR) wavelength. However, since the human vision is sensitive only in the visible region, a hot mirror/infrared blocking filter is used in cameras. Certain complimentary attributes of NIR data are, therefore, lost in this process of image acquisition. However, RGB and NIR images are captured entirely in two different spectra/wavelengths; thus they retain different information. Since NIR and RGB images compromise complimentary information, we believe that this can be exploited for extracting better features, localization of objects of interest and in multi-modal fusion. In this paper, an attempt is made to estimate the NIR image from a given optical image. Using a normal optical camera and based on the compressed sensing framework, the NIR data estimation is formulated as an image recovery problem.