Xiangyin Zhang

An approach of point cloud denoising based on improved bilateral filtering

An omnidirectional mobile platform is designed for building point cloud based on an improved filtering algorithm which is employed to handle the depth image. First, the mobile platform can move flexibly and the control interface is convenient to control. Then, because the traditional bilateral filtering algorithm is time-consuming and inefficient, a novel method is proposed which called local bilateral filtering (LBF). LBF is applied to process depth image obtained by the Kinect sensor. The results show that the effect of removing noise is improved comparing with the bilateral filtering. In the condition of off-line, the color images and processed images are used to build point clouds. Finally, experimental results demonstrate that our method improves the speed of processing time of depth image and the effect of point cloud which has been built.

Semantic Segmentation Based on Deep Convolution Neural Network

Semantic segmentation using full convolutional neural network (FCN) avoids the problems of repeated calculation and storage due to using of pixel blocks. However, the results obtained by FCN are still not precise enough, and the results of upsampling are still relatively fuzzy and smooth. It is not sensitive to the details such as small object in the image. Therefore, this paper proposes an image segmentation method based on simplified deep residual network. A simplified Deep Residual Network (DRN) is proposed to replace the VGG-16 network in the original FCN framework. The net combines deep residual network architecture with 30% lesser parameters than the VGG model and the method of skip connection that combines semantic information from a deep, coarse layer with appearance information from a shallow, fine layer to produce accurate and detailed segmentation results. The results showed that the effect of segmentation is more refined and the recognition and segmentation of small objects are significantly improved by the improved network model.

A novel phase angle-encoded fruit fly optimization algorithm with mutation adaptation mechanism applied to UAV path planning

Abstract This paper proposed an improved version of fruit fly optimization (FOA) to solve the unmanned aerial vehicle (UAV) path planning problem. The improved algorithm combines the phase angle-encoded and mutation adaptation mechanisms into the basic FOA and is referred to as θ-MAFOA. Mutation adaptation mechanism is adopted to enhance the balance of FOA in terms of the exploitation and exploration ability, while phase angle-based encoded strategy for fruit fly locations helps to achieve the high performance in the convergence process. Then, the proposed θ-MAFOA is used to find the optimal flight path for UAV in the complex three-dimensional (3D) terrain environments with various ground defense weapons. B-Spline curve that connects the start and target points is employed to represent a smooth path. Several performance criteria and constraints are taken into consideration to evaluate the cost of UAV paths. Numerical experiments are carried out on various test scenarios and the results show the proposed θ-MAFOA is superior to the basic and other two modified versions of FOA, and also more powerful than several existing state-of-the-art optimization algorithms.