Image, Vision, Computer_Problems Faced by Computer Vision Image Measurement Research

People have carried out fruitful research on the above-mentioned research contents, and developed the technology and algorithm of the big one, and they have been widely used in various fields. However, most of the various visual image measurement systems established by people at present are only applicable to a certain environment or application occasion, and it is necessary to establish a general vision system comparable to the human visual image measurement system. It is very difficult, and the main reasons are reflected in the following aspects. 1) The three-dimensional scene of image ambiguity is projected as a two-dimensional image. The depth information and the invisible part of the information are lost. Therefore, an unshaped one-dimensional object will be projected on the image (the problem of the same image on the image plane). In addition, the images of the same object obtained from different angles will be very different, as shown in Figure 1-6. Therefore, additional constraints are needed to solve the ambiguity when recovering the scene from the image. The multi-dimensional image visual image Definition 2) Environmental factors affect many factors in the scene, including background lighting, light source angle, object shape, camera and spatial relationship changes, air conditions, and surface color will all have an impact on the projected image. When any... When the organs change, it will have an impact on Pianxu. And, all these factors are attributed to a single measurement result, that is, the gray scale of the map. It is very difficult to determine the effect and size of various factors on the gray scale. 3) Understanding natural scenes requires a lot of knowledge. Understanding natural scenes requires a lot of knowledge, for example, knowledge of shadows, textures, stereo vision, and object size. Specialized knowledge or general knowledge about objects, and possibly relationships between objects. Due to the large amount of knowledge required, it is difficult to simply input manually, and may have to be established through automatic knowledge acquisition methods. 4) The amount of data is large. Gray scale images, color images, and depth maps. Your information is very large. Huge data like requires a lot of storage space, and it is not easy to realize fast processing. For example, the data volume of a grayscale image with a resolution of 512×512 is 256KB, and the data volume of a color image with a resolution of 1 division is 768KB. If the image sequence is processed, the data volume will be larger. This is looking forward to new breakthroughs in high-speed array processing units and algorithms (such as neural networks, fractal dimensional algorithms, wavelet transforms, etc.), with very little calculation and high parallelism to achieve the function of processing large amounts of data. In order to solve the problems faced by computer vision, researchers continue to seek new ways and means, for example, active vision, task-oriented vision (task-oriented vision), knowledge-based, model-based vision, and multi-sensor fusion and integration Vision and other methods, among which people are paying more and more attention to the application of knowledge. As we will see, the biggest feature of a computer vision system is that in all stages of vision, the system performs as much automatic calculations as possible. To this end, the system needs to use a variety of knowledge, including feature models, adult processes, object models, and relationships between objects. If the computer vision system does not make use of this knowledge, the scope of its application and its functions will be limited. Therefore, the visual system should use the knowledge that can be clearly expressed, so that the system has higher adaptability and robustness. Reasonable use of knowledge can not only effectively improve the adaptability and robustness of the system, but also solve difficult problems in computer vision. The human visual system has high-resolution characteristics, and has three-dimensional observation, superior recognition ability and flexible reasoning ability. The complex and profound visual mechanism makes some scholars can't help but lament that if it weren't for the existence of the human visual system as an example of the universal visual system, the tomb would doubt whether they could find a way to establish a universal visual system. Because of this, endowing machines with human visual functions has been a goal that people have been pursuing unremittingly for several decades.