A virtual campus roaming system is established base on the Fushun University. Two kinds of key technologies are Focused on, one is the methods of 3D modeling and optimization based on architectural working drawing, the other is the dynamic interaction between the user and the virtual environment on campus. The virtual campus is d eveloped in AutoCAD and 3DS MAX. The final virtual campus system provides roaming and collision detection function independently through the Virtools. The virtual campus has been carried out now and is of good effect．

To advance the visibility of the fog-degraded image, a method for fog-degraded image clearness was presented. To restrain the noise amplification in sky region，the sky region first was segmented before enhancement, then using optimal normal distribution to put the pixels of the high grey region into that of the fuzzy，the range of pixel value in the sky can be got. By the information，the sky region can be segmented from the image. A moving mask was adopted to segment scene at different depth，then each pixel in the mask is processed using block-overlapped bi-histogram equalization. The experimental results show that this method can efficiently improve the degradation of image and enhance the clearness of image.

The image processing effect based on morphology is depending on the operation and the structural element. To suppress noise of image and withdraws the edge information. A method which uses multi-element composite structure with dual gradient filter and multiple synthesis methods based on soft morphology was presented. Flexible multi-element composite structure could effectively remove the image noise filtering, and dual-gradient multiplex could improve the detection rate of the image edge. The experimental results show that the method is better than traditional algorithms in suppressing noise and edge detection.

The moving object trace detection based on sequence images is a critical technology of target recognition and track processing. The core of the motion trajectory detection is the contour character extraction of the moving object, however traditional algorithms only highlights partial characters of the contour because of the different factors. On the basis of extracting different characteristic information from traditional algorithms, the multi-template integration edge detection method of ant colony clustering was adopted to obtain the edge of moving object. The experimental results show that this method can obtain relatively complete characteristic information in the course of trace detection and can attain the trace of the object more effectively.

Blink is an important feature of human eyes. The both eyes were treated as an entirety and the eye area was located by using Haar wavelet. The two eyes were obtained based on facial model, at last blink detection method were got by using PCA method. The method reduced error of eye location using single eye by treating the both eyes as a whole. And based on the result of eyes detection data extraction ,blink detection can be done by using PCA method. The experimental results show the method is feasible and effective.

The core technology of reverse engineering is free surfaces reconstruction. Traditionally , surface reconstruction is built directly by using reconstruction data , which control vertexes aren't compressed. So it has a relatively weak data processing ability , and the storage and calculation of data in the reconstruction process is tremendous. A method applying wavelet transform technology , which can reduce the amount of controlling vertexes of cubic quasi - uniform B - spline surfaces, was presented. The amount of control vertexes to reconstruct curved surfaces decreases greatly . The curved surface with its data compressed approximates the original surface shape. Experiment results show that the method implements surfaces'hierarchical compressing , simplifies data processing and reduces the cost of reconstruction and accords with engineering demand.

    Based on mathematical morphology of multi -structure element , an algorithm of image edge -detection was pro posed. With edge threshold and noise having different morphology , while detecting the edge , it needs determining the pixel whose gradient varies obviously .Then it applies multi -structure element to erosion operating of binary morphology to determine that the pixel is edge or noise.If it is noise , the pixel is removed, otherwise kept .In order to test the effect of algorithm , the results of this algorithm and of several traditional algorithms to extract edge of image were given .Experiment results show that the algorithm detects image edge successfully , and effect of edge -detection is evidently better than classical Sobel operator and Canny operator .

        As a new technique applied to protecting the copy right of digital productions, the digital watermark technique has drawn extensive attentions.The watermark technique based on wavelet transform has been an active study field .In order to ensure the concealment and robustness of watermark, a blind image algorithm on the basis of wavelet transform was proposed.Firstly ,watermark was scrambled before the algorithm w as embedded .Then watermark information and spread spectrum sequence are modulated, and combining texture concealable characteristics of human vision system with just differenthating limit, were embedded into DWT domain of carrier image .Finally , making use of autocorrelation function characteristic of spread spectrum sequence ,watermark information was detected .In the detection process, it is not requisite for initial carrier image .The experiment results demonstrate that with the algorithm embedded watermark, watermark is no t easy to be found out .I t has a fine detected watermark effect .The algorithm is robust enough for JPEG degradation compression, clipping , noise attacks and low -pass filtering .