Development of Technique for Recognition of the Complex Shape Signal Images

Abstract

With the help of digital image conversion, they create conditions for improving the perception of the image, the formation of a certain artistic image, the allocation of informative features. Image conversion is carried out by various methods, including optical, photochemical and digital methods. The latter are becoming increasingly common, as technical capabilities only grow from year to year. The purpose and objectives of this work: to analyze the main methods of image processing and recognition; to develop an effective algorithm for recognizing signal images; to develop a software product that conducts experiments to recognize complex waveforms. Information about the degree of similarity of the two signals is shown at their maximum coincidence and is determined by the behavior of the correlation integral (functional) at the extreme point, that is, the value of the integral and its derivatives. This functional is an energy characteristic, the value of which is determined by the brightness and area of the image visible through the standard. Its extremum is observed in cases where the standard is fully fit in the image, and the coincidence of their shape is not necessary. Therefore, this value characterizes the similarity of the image and the standard in form ambiguously. The absolute value of the second derivative of the correlation integral to the point of its extremum is preferable for estimating the similarity of the image and the standard in form. The dependence of the functionals on the degree of "blur"is presented. A necessary condition for the practical value of such recognition algorithm is the independence or weak dependence of the functional threshold value on the shape, area and contrast of the image. The table shows the values for some images with different area and contrast. as an indicator of the similarity of the image and the standard in form, it is advisable to use the module of the second derivative of the correlation integral normalized in contrast at the extreme point. First, the program receives the original image in bmp or jpeg format, and shows it to the user. The image is then muted according to the noise type selected by the user and the parameters set for the selected noise type, then displayed on the form. You can save an image, run an experiment (plot), define a shape, or noise the image again. To compare the capabilities of the algorithm under study and the human visual system according to the classification of "blurred" images, the images of simple geometric shapes are shown recognized by an automatic device implemented in the form of a software tool. The type of distortion that can be used: monochrome noise by Gauss, color noise by Gauss, spray, turn into mosaic, blur. The program compares the distorted image with each of the available reference figures, estimates the total standard deviation, and concludes what image is depicted, based on its minimum value.