Research content, image, vision_The research content of computer image vision
Next, the main research content of computer image vision is introduced from five aspects: input equipment, low-level image vision, middle-level image vision, high-level image vision and architecture. Computer image measurement research guide Figure 1) Input device Input device (input device) includes security equipment and digital equipment. Adult equipment refers to the detection of surrounding scenes or objects through optical camera or infrared, laser, ultrasound, X-ray, and then use digital equipment to obtain two-dimensional or three-dimensional digital icons about the scene or object. Obtaining digital images is the most basic function of the computer imaging vision system. At present, most of the input devices used for image vision research are commercial products, such as CCD black-and-white or color wing cameras, digital scanners, ultrasonic adult detectors, CT (computed tomography) security equipment, etc. However, these commercial input devices are far from meeting actual needs. Therefore, many researchers are still studying various advanced performance systems, such as infrared positioning systems, laser beaming systems, and adjusted calculations. Computational imaging system, that is, each exclamation element (or several generous elements) corresponds to a simple processor, so that it can adapt to situations where complex scenes change dynamically. 2) Low-level image vision Low-level image vision (low level vision) mainly processes the input original image. This process borrows a large number of image processing techniques and algorithms, such as image filtering, image enhancement, edge detection, etc., in order to extract the basic features of the scene such as corners, edges, lines, borders, and colors from the image. This process also includes various image transformations (such as correction), image texture detection, image motion detection and so on. 3) Middle level image vision The main task of middle level vision is to restore the scene's depth, surface normal direction, contour and other 2.5-dimensional information about the scene. The realization methods include stereo vision and distance measurement. Ranw finder, motion estimation, and the use of bright W, features, texture features and other methods for shape recovery. Research contents such as system calibration and system adult model are generally also carried out at this level. 4) High-level image vision The main task of high-level vision is to restore the complete three-dimensional image of the object based on the original input image, basic characteristics of the image, and 2.5-dimensional image in the object-centric coordinate system. , Establish a three-dimensional description of the object, recognize the three-dimensional object and determine the position and direction of the object. In addition, active vision covers the above-mentioned research content at all levels. In CT and SAR (Synthetic Aperture Radar), the processing method of Tukan reconstruction is used. It is worth pointing out that low-level image vision, middle-level image vision, and high-level image vision basically correspond to the three stages of Marr image vision. 5) System architecture The most common meaning of the two terms system architecture refers to the study of the structure of the system at a high level of abstraction, based on the system model rather than the specific examples of the implementation design. To illustrate this point, one can consider the difference between the architectural style of a certain period of time (such as the Qing Dynasty) in architectural design and the specific buildings designed according to these two styles. Architecture research involves two series of related topics, including parallel structure, hierarchical structure, information flow structure, topological structure, and the approach from design to implementation.