Spatil Depth Detection Based On Structured Light And Reserch On3D Reconstruction

Structured-light measurement is not only an effective means of non-contact measurement, but also an actual application of computer vision theory in practice. A growing number of structured-light measurement products are applied in our daily life and production for the time being, which are supported by a theoretical system perfectly. The structured-light measurement has a good performance in accuracy and rapidity, and the efficient encoding ways can also make the measurement have a good dynamic character. In many practical applications of computer vision theory, structured-light measurement has long been the focus of research and application.Triangulation measurement principle is a classical measuring principle and widely used in many different fields such as positioning, space-measuring and detection. Triangulation is a traditional common measurement method and utilized in various situations. The structured-light measurement is also based on the basic principle of triangulation, but mostly just used for simple distance measurement in the early days of its emergence. After nearly ten years of development, many different acquisition and measuring methods were proposed for structured-light measurement, including linear laser measurement, multi-line laser measurement and coded-light measurement in addition to laser ranging. With the improvement of those methods, structured-light measurement emerges as a system and is applied in more applications nowadays.Beginning with the coded structured-light, a pseudo random code array(M-sequence) with good window-characteristics is produced by means of simulating the computing manner of linear feedback shift register (LFSR), so as to make the projected light on the measured object turn into discrete faculae. According to the faculae shapes and locations gathered with a camera, the tasks such as decoding, positioning and calculation of the depth information are subsequently performed. A measurement space ranging from1000mm to5000mm is therefore constructed, and the measurement accuracy is controlled within10mm.In the coded structured-light measurement process, images with a large number of discrete symbols are acquired. The direct use of traditional threshold segmentation method for extracting and recognizing may cause the loss of the symbolic information. In view of this, the traditional method is not suitable for global threshold segmentation. The structured-light image contains a great deal of discrete connected domains and the very arrangement of these domains constitutes a form of certain image contents. In order to distinguish the form, it is necessary to separate and identify all symbols in the region of discrete symbols. Obviously, traditional methods cannot solve the problem effectively because most of the common segmentations can only treat a single connected domain. Aiming at overcoming the shortcomings of traditional segmentation methods, a new segmentation algorithm based on the principle of mathematical statistics was presented. The algorithm qualitatively outlined the whole range of discrete symbols by analyzing the light density, so as to distinguish the foreground and background of the image. Experiments showed that the parameter-controlled contours of certain regions can be easily obtained with this algorithm and therefore the foreground can be extracted and segmented from the image with the contours available, which is conducive to improve the results of the following processes such as image processing and character recognition. The experimental results demonstrated the validity of the proposed algorithm.In this method, a program was designed to analyze and process the measurement results, and a specific data structure was designed for storing depth information. The measurement results were preserved using depth-map and points cloud, and a group of16bit pixel data storage units were constructed according to the measuring accuracy and range to save the depth information which is represented by each pixel. To visually demonstrate the measurement results, OpenGL and other tools were used to draw an image of the points cloud and render the reconstructed3D information of the measured object in the computer.