In order to solve the problem that static attribute reduction cannot update attribute reduction efficiently when the number of attributes in the set-valued decision information system changes continuously， an incremental attribute reduction method with knowledge granularity as heuristic information was proposed. Firstly， the related concepts of the set-valued decision information system were introduced， then the definition of knowledge granularity was introduced， and its matrix representation method was extended to this system. Secondly， the update mechanism of incremental reduction was analyzed， and an incremental attribute reduction method was designed on the basis of knowledge granularity. Finally， three different datasets were selected for the experiments. When the number of attributes of the three datasets increased from 20% to 100%， the reduction time of the traditional non-incremental method was 54.84 s， 108.01 s， and 565.93 s respectively， and the reduction time of the incremental method was 7.57 s， 4.85 s， and 50.39 s respectively. Experimental results demonstrate that the proposed incremental method is more faster than the non-incremental method under the condition that the accuracy of attribute reduction is not affected.

Aiming at the problem that the performance of parallel computing cannot be improved by extending its scale under the constraint of fixed structure, a method of proportionally adjusting graph weights was proposed to handle such extension problem. The method firstly investigated the factors from architecture and parallel tasks which affected its scalability, and then modeled the architecture as well as parallel tasks by using weighted graph. Finally, it realized an extension in parallel computing by adjusting proportionally the weights of the vertices and edges in the graph model for parallel computing. The experimental results show that the proposed extension method can realize isospeed-efficiency extension for parallel computing under the constraint of fixed structure.

An improved Canny operator was proposed to ensure the accuracy of the Light Emitting Diode (LED) wafer edge detection applied in the vision system of automatic LED die bonder. In the traditional Non-Maxima Suppression (NMS) process, inaccuracy would be caused by selecting two pixels near gradient direction as contrastive points. To solve this problem, bilinear interpolation based on the center pixel and three pixels around was made along gradient direction to implement new NMS process. In addition, the gray histogram of LED wafer was characterized by typical three peaks, and then the high threshold and the low threshold could be got by Otsu dual threshold method with new evaluation function to replace traditional artificial adjustment. The experimental results indicate that greater wafer outline and poles edges can be got by the improved Canny algorithm which is appropriate for LED wafer edge detection.

Elliptic Curve Cryptosystem (ECC) is a kind of public-key cryptosystem based on algebraic curve. Point algorithms on elliptic curve are essential operations in ECC and its speed are decided by selected coordinate. On the basis of analyzing existing coordinates such as affine coordinate, Standard projective coordinate, Jacobian projective coordinate and Lopez & Dahab projective coordinate, an extension of Lopez & Dahab projective coordinate was proposed and a mixed projective coordinate was given to improve the speed of scalar multiplication. The result of analysis indicates that the operation time complexity of scalar multiplication under the mixed projective coordinate is more reduced than under any other coordinates.