Search

This paper improves the Consensus-Based Bundle Algorithms (CBBA), and solves the problem of heterogeneous Unmanned Aerial Vehicle (UAV) complex task allocation under timing constraints. In order to solve the task assignment problems with constraints such as the heterogeneity and resource constraints of the UAV, the diversity of tasks, and timing constraints, a dynamic task generation mechanism is proposed based on the CBBA. According to the constraints, the complex task is decomposed into multiple simple tasks for assignment, and the problem is converted into a single UAV task assignment problem for resolution. The simulation results show that based on satisfying the constraints, the improved CBBA algorithm can achieve the balance between maximizing the total objective function and minimizing the calculation time, and find the optimal solution for task allocation.

Edge detecting is the most important pre-processing in image analysis, feature extraction and recognition. Gradient is a basilic measure of edge images, and we can calculate the values of magnitude and rotation of edges with it. We analyse the first-order difference detectors first, such as Robert, Prewitt and Sobel detectors, then we analyse the Laplacian detector, that is a second-order difference one. A generalize mask is found based on the relationship of the classical edge detectors, and different value of its parameter is correspond to the detectors. The experiment show that the edges detected with the generalize mask are same to those detected with classical detectors, so the mask is reasonable.

This paper proposes that box-counting dimension method be adopted for the analysis of remote sensing images of Chinese mountains. At first, remote sensing images of half-bared and green Chinese mountains are segmented with a global image threshold value, then the binary images are calculated with Box-counting method. The authors find that the distributions of ln(N(l)) and ln(l) are almost stand on a line, which indicate that the fractal characters of the images are obvious, especially when the value of ln(l) is greater than -2; the width of band of distributions of ln(N(l)) and ln(l) of green mountain images is wider than that of half-bared ones, but this has not influence of the value of FD of both kinds of Chinese mountains.