An image information extraction algorithm on fractional differentials is put forward in this paper that is based on the characteristics of fractional differential in signal processing. This paper has extracted the information of salt and pepper noise images with various coefficients, and analyzed and compared it with the information extraction results of classic integer-order operators as Prewitt, Roberts and Sobel. Experiments have shown that not only the high-frequency marginal information can be extracted by extracting information with fractional differentials, just as it is extracted with integer-order operators, but the texture information can also be extracted from the smooth region. Besides, this algorithm is featured with great noise immunity against salt and pepper noises.

Texture and edge information are equally important to identification of most of natural images. To the problem of the existing integer order differential operators can’t extract texture information form the images, this study developed a fractional differential algorithm, which can extract texture and marginal information simultaneously, based on settlement of the drift problem of fractional differential operator. Experimental results showed that our algorithm can not only extract texture information but also extract more edge information than the traditional algorithms. And to the image with Gauss noise, our algorithm also have noise immunity.

By comparing and analyzing various orders coefficient curves of fractional order, this article studies the influence of order to the extraction of signal memory information, and then extends the result to image processing, as well as analyzes the influence of fractional order differential order to texture information extraction, then gives the order scope of texture information extraction.

Aiming at the fatal flaws of the traditional diagnosis methods for the large-scale photoelectric tracking devices, such as poor stability and adaptive capacity, lack of inspiration and narrow domain knowledge of expert system, etc, more importantly, fundamentally improve the diagnostic efficiency and universality, in this paper, an intelligent mixed inference diagnosis expert system based on multiple knowledge representation and BP neural network is put forward. Firstly, some related key basic concepts and principles of intelligent fault diagnosis technology and several major applied diagnosis knowledge representation methods such as diagnosis fault tree, frame representation production rule and so on, were elaborated. Secondly, in view of high concurrency and relevancy of the system faults, a mixed reasoning mechanism combining BPNN and ES was researched. Finally, some interrelated essential implementation techniques, such as system architecture and VR technology, were also presented. Actual applications and experiments demonstrate that the proposed approach is robust and effective.