Content-based Image Retrieval Technology With Application In Military Reconnaissance

With the development of Modern Technology especially High Tech, skills and capabilities concerning Military Inspection have made great progress, too. Thus, it's of vital importance to acquire information required out of vast amount of images derived from Military Inspection. To meet such need of analyzing images, we have to target at those valuable and worthy ones, and then recognize specific objects described by related graphs we get. The process itself invites techniques of both image segmentation and feature matching. So far, world wide application of image segmentation has been mostly used in various non-ballistic missile tracking devices with common thermal infrared image tracking or terrain matching guidance techniques. However, detailed technological methods are not revealed to the public. In China, reports in military inspection enjoy a wide popularity, yet techniques of image segmentation and feature matching in target recognition remains a new research field worth exploring. Thus, this paper emphasizes on tanks, employs image segmentation and matching techniques, and identifies military targets in the following discussion.Techniques gaining major research efforts include image smoothing, medium value filtering and image sharpening. Besides, two kinds of image segmentation techniques are mentioned here, i.e. histogram-based segmentation and marginal extraction algorithm with bettered active-outline model. The author analyzes target feature selection principles and methods based on target feature matching algorithm and refers to outline feature LCS description approach raised by Lalit Gupta et al. Finally, he proposes a target outline feature description method, Target Outline Sequence TLMCS description, with abilities of both outline description and local part inspection. Then, with the support of maximum local outline sequence extraction, an algorithm dealing with target feature extraction sorting is designed, based on the previous TLMCS. This algorithm carries out pretreatment of ordinary military graphs, then segments images, extract edges in the margin, and finally conducts feature matching. Experiments results show that pretreatment technique employed in this paper not only reduces disturbances from noises and fake images, but also improves segmentation precision. Among them, the improved initiative profile mode draw-out algorithm not only can precisely drawing out in the image the convexity object edge;but also can approach the hollow place of the edge。At the same time, It introduced the auto-adapted changing size exterior restraint energy to increase the attraction scope, enables the control point not to...