Detection and classification of road markings are a prerequisite for operating autonomous vehicles. Although most studies have focused on the detection of road lane markings, the detection and classification of other road markings, such as arrows and bike markings, have not received much attention. Therefore, we propose a detection and classification method for various types of arrow markings and bike markings on the road in various complex environments using a one-stage deep convolutional neural network (CNN), called RetinaNet. We tested the proposed method in complex road scenarios with three open datasets captured by visible light camera sensors, namely the Malaga urban dataset, the Cambridge dataset, and the Daimler dataset on both a desktop computer and an NVIDIA Jetson TX2 embedded system. Experimental results obtained using the three open databases showed that the proposed RetinaNet-based method outperformed other methods for detection and classification of road markings in terms of both accuracy and processing time.

In order to capture an eye image of high quality in a gaze-tracking camera, an auto-focusing mechanism is used, which requires accurate focus assessment. Although there has been previous research on focus assessment in the spatial or wavelet domains, there are few previous studies that combine all of the methods of spatial and wavelet domains. Since all of the previous focus assessments in the spatial or wavelet domain methods have disadvantages, such as being affected by illumination variation, etc., we propose a new focus assessment method by combining the spatial and wavelet domain methods for the gaze-tracking camera. This research is novel in the following three ways, in comparison with the previous methods. First, the proposed focus assessment method combines the advantages of spatial and wavelet domain methods by using ε-support vector regression (SVR) with a symmetrical Gaussian radial basis function (RBF) kernel. In order to prevent the focus score from being affected by a change in image brightness, both linear and nonlinear normalizations are adopted in the focus score calculation. Second, based on the camera optics, we mathematically prove the reason for the increase in the focus score in the case of daytime images or a brighter illuminator compared to nighttime images or a darker illuminator. Third, we propose a new criterion to compare the accuracies of the focus measurement methods. This criterion is based on the ratio of relative overlapping amount (standard deviation of focus score) between two adjacent positions along the Z-axis to the entire range of focus score variety between these two points. Experimental results showed that the proposed method outperforms other methods.