We present a self-supervised depth estimation approach using a unified volumetric feature fusion for surround-view images. Given a set of surround-view images, our method constructs a volumetric feature map by extracting image feature maps from surround-view images and fuse the feature maps into a shared, unified 3D voxel space. The volumetric feature map then can be used for estimating a depth map at each surround view by projecting it into an image coordinate. A volumetric feature contains 3D information at its local voxel coordinate; thus our method can also synthesize a depth map at arbitrary rotated viewpoints by projecting the volumetric feature map into the target viewpoints. Furthermore, assuming static camera extrinsics in the multi-camera system, we propose to estimate a canonical camera motion from the volumetric feature map. Our method leverages 3D spatio- temporal context to learn metric-scale depth and the canonical camera motion in a self-supervised manner. Our method outperforms the prior arts on DDAD and nuScenes datasets, especially estimating more accurate metric-scale depth and consistent depth between neighboring views.

Change/motion detection is a challenging problem in video analysis and surveillance system. Recently, the state-of-the-art methods using the sample-based background model have demonstrated astonishing results with this problem. However, they are ineffective in the dynamic scenes that contain complex motion patterns. In this paper, we introduce a novel data-driven approach that combines the sample-based background model with a feature extractor obtained by training a triplet network. We construct the network by three identical convolutional neural networks, each of which is called a motion feature network. Our network can automatically learn motion patterns from small image patches and transform input images of any size into feature embeddings for high-level representations. The sample-based background model of each pixel is then employed by using the color information and the extracted feature embeddings. We also propose an approach to generate triplet examples from CDNet 2014 for training our network model from scratch. The offline trained network can be used on the fly without re-training on any video sequence before each execution. Therefore, it is feasible for real-time surveillance systems. In this paper, we show that our method outperforms the other state-of-the-art methods on CDNet 2014 and other benchmarks (BMC and Wallflower).

The rectification process is a compulsory step in stereo matching computation. To obtain depth information, stereo camera systems are often installed in vehicles for outdoor and street-related applications, including vehicle and pedestrian detection, lane detection, and traffic sign recognition. In this paper, we propose a rectification method that uses currently available front- and rear-view vehicle cameras to produce rectified stereo images. The proposed method can be employed with different types of cameras that have varying focal lengths. In addition, this method tolerates the problem of camera alignment variation from normal stereo camera systems. To achieve this, a compensation method for different focal lengths and the estimation of image relationships are introduced. The experimental results demonstrate that the proposed method can operate robustly and accurately with different kinds of stereo images and significantly outperforms a state-of-the-art rectification method.

Background subtraction has been a fundamental task in video analytics and smart surveillance applications. In the field of background subtraction, Gaussian mixture model is a canonical model for many other methods. However, the unconscious learning of this model often leads to erroneous motion detection under high variation scenes. This article proposes a new method that incorporates entropy estimation and a removal framework into the Gaussian mixture model to improve the performance of background subtraction. Firstly, entropy information is computed for each pixel of a frame to classify frames into silent or high variation categories. Secondly, the removal framework is used to determine which frames from the background subtraction process are updated. The proposed method produces precise results with fast execution time, which are two critical factors in surveillance systems for more advanced tasks. The authors used two publicly available test sequences from the 2014 Change Detection and Scene background modelling data sets and internally collected data sets of scenes with dense traffic.