3D maze has reached a high level in terms of complexity and fun. By improving the algorithm of generating a 2D maze, we propose the concept of loop maze and the complexity formula of the maze. Then, an algorithm for designing a 3D maze is presented based on the quadrilateral mesh surface. This approach mainly consists of three steps:Firstly, the quadrilateral mesh is generated on the given 3D surface; Secondly, the start point and end point of the maze are chosen alternatively, and the maze on the quadrilateral mesh surface is obtained by the algorithm of generating a 2D maze which is based on a spanning tree algorithm; At last, the maze is turned into 3D structure, and 3D maze is generated by Boolean operation between the 3D structure and the original 3D model. Several personalized 3D maze toys are produced by 3D printer, which consumedly enhances fun and user experience.

Based on the data of China's industrial enterprises above the designated size, this study adopts several spatial statistics methods, such as the standard deviation ellipse, average nearest neighbor analysis and kernel density estimation to measure the spatial distribution pattern of private enterprises investing in the United States in Zhejiang province in 2006, 2009, 2012 and 2015 and applies the geographic probe to analyze the geographical factors which affect the spatial distribution of these enterprises. The results show that the location of these private enterprises agglomerate along the long axis of the standard deviation ellipse and diffuse along the short axis, while the distribution center is always moving to the northwest, and always in Shaoxing city; the nearest neighbor index of these enterprises is continuously reduced; the high-density region is mainly distributed in the developed region in the northeast of Zhejiang province, and is centrally distributed with Hangzhou urban area, Ningbo urban area and "Cixi - Yuyao" as the core, followed by the urban area of Wenzhou, while a few private enterprises is distributed in the rest of the economically less developed area; The factors influencing the spatial distribution of these enterprises turn gradually from a single factor to multiple factors, among them, the economic development level, technological innovation capability and foreign direct investment are factors which have the strongest influence.

Taxi, as an important supplement of urban public transport, plays an important role in resident daily traffic. However, because of the unbalanced spatial and temporal distribution of the potential passengers, high no-load rate and unhealthy competition keep releasing negative impacts on taxi business. With the development of ITS (intelligent transport system) technology, more and more attempts have been made to study approaches of seeking taxi passengers, while neglecting the analysis of high profitable seeking strategy. Based on mathematical statistics and geo-statistics, we propose a novel spatial and temporal analysis method based on an evaluation index, which takes the combination of income per unit time in the load state and the time for seeking passengers in the adjacent no-load state. We investigated the Shenzhen's taxi data set, which contained 13 798 taxis GPS trajectories during a week from April 18 to April 24, 2011 to conduct a comprehensive analysis on the characteristics of taxi operation and the spatial and temporal distribution of people's daily traffic. Our study indicates that the designed method yields a better reflection on the distribution of high profitable passengers, providing a new perspective for the efficient operation of the taxi.

Medical image data are rapidly accumulating and traditional image analysis methods based on manual approaches has imposed a heavy burden on doctors. Computer vision has played an important role in alleviating the pressure of manual reading,improving the accuracy of diagnosis and promoting the standardization of medical procedures by providing automatic or semi-automatic auxiliary diagnostic methods. At present,deep learning convolutional neural network has achieved outstanding performance in various medical image processing tasks,but the unexplainability of deep learning "black box" has become a major obstacle to further explore the full potentials of intelligent medical diagnosis.This paper summarizes the research progress of deep learning interpretability in medical image processing in recent years.Firstly,we clarify the application status and problems of deep learning in the medical field,and discusses the interpretable connotation of neural networks. Then,we focus on the research progress of deep learning interpretability in medical image data processing starting from the common methods of deep learning interpretability. Finally,the interpretable development trend of medical image processing is discussed.

Lung cancer is one of the most aggressive cancers and detecting lung nodule by CT images at the early stage is of vital importance to treating lung cancer. This paper overviews the application of a revolutionary image recognition method, deep learning, in the detection of lung nodule. First, we contrast different convolutional neural network (CNN) architectures and their performance in image recognition. Then, we mainly focus on various deep learning methods including faster-RCNN, transfer learning, residual network and curriculum learning to train the classifier. We also introduce some available databases of lung CT images in the last section of our paper.