Search

Existing route recommendation systems have three main weakness. Firstly, they usually assume that there is always an alternative sub-route during the whole driving process without considering whether there is a substitute or not when the traffic condition changes. Secondly, they confuse the optimal road with shortest road. The optimal road differs from the shortest road that always means the road with shortest time or shortest distance. Thirdly, they haven't taken full advantage of historical data. Most of time they only use historical data to compute average speed of each road. In this paper, we focus on these problems especially on how to choose a substitute route. We come up with a new idea to recommend routes by considering replaceable routes. It takes whether the road is replaceable or not into account to actually recommend reliable route and ease cities' traffic stress.

According to the nonlinear characteristics of original sonar image with big data, adaptive narrowband level set model to detect sonar image is proposed in this paper. In order to estimate the approximate position of the underwater objects after smoothing the original image, initial segmentation is processed with the block mode k-means clustering algorithm. On this basis, zero level set function is adaptively initialized using the approximate position of the underwater objects, which can not only reduce human intervention but also improve the detection speed. To further improve the robustness and detection accuracy, new adaptive narrowband level set model are provided to complete local optimization by minimizing each new energy function. Detection experiments demonstrate that the proposed method improves the detection speed and accuracy, and it has certain adaptability.

To reduce the tracking errors caused by the background change and occlusion in dynamic scenes, a novel Mean-Shift moving target tracking algorithm based on weighted sub-block is proposed in this paper. Moving target tracking is completed by blocking the target region. The weight of each sub-block is determined by the combination of the similarity between target sub-block and candidate sub-block and the ratio of the target sub-block area and the overall area. At the same time, the edge position of the target sub-block is found by means of a Sobel operator edge detection algorithm. By which, the target sub-block area is obtained. Both of the target region's RGB color information and the pixel's position information are taken into consideration while describing the characteristic model of target and candidate region inside each sub-block. The experiments demonstrate that the proposed method is insensitive to the background change and occlusion, and has better tracking performance with higher tracking accuracy and adaptability.