The application of morphological algorithms for the restoration and segmentation of graphic data for the machine vision system

Abstract

An analysis of current methods of machine analysis of graphic data, used within the organization of the machine vision system and based on both software and neural network algorithms, was carried out. The effectiveness of the application of methods of preprocessing graphic data by software algorithms, which includes the application of image restoration and segmentation operations, is indicated, and provides an opportunity to both reduce the load on the computing resource of the hardware and software platform and reduce the time of machine analysis of streaming data. The peculiarities of the application of morphological methods of image preprocessing, which include the procedures of erosion, dilatation and selection of borders when working with graphic data presented according to the RGB color scheme, are indicated, and a generalization of the specified technique for working with alternative color schemes is carried out. In a similar way, an analysis of the features of the adaptation of the method of determining the image matching in accordance with the color scheme of the representation of graphic data was carried out; it is shown that the mentioned approach in combination with morphological methods can be used both at the stage of pre-processing and at the stage of image segmentation. Also, the complex methodology presented in the study included the application of a threshold algorithm, which consisted of the initial assessment of the threshold value and the accuracy of its calculation, as well as the cyclic performance of the operations of dividing the basic elements of the image into areas with the calculation of the average values of their parameters and the correction of the threshold value. The final stage of segmentation included the application of a watershed algorithm adapted to the color scheme in relation to a set of basic image elements, which includes the expansion of segments through the sequential application of morphological algorithms. According to the research results, the problem of excessive segmentation is effectively solved at this stage through the use of pre-processing and, in particular, the selection of image matching. In this way, a complex methodology for the restoration and segmentation of graphic data was built based on software algorithms, the settings and optimization of which can be carried out in automatic mode according to the target indicators of the efficiency of restoration and segmentation, the ratio of the time of machine analysis in relation to the volume of input data, as well as load on the computing resource.