Optimization of Identification of Micro-Objects with Blurring of Image Points

Scientific and methodological foundations have been developed for the identification and recognition of micro-objects in the systems of environmental protection and ecology, medicine, aimed at using redundant information structures of image components - histological, morphological, fractal, geometric, and other characteristics. The effectiveness of the mechanisms for identifying micro-objects according to the criteria of the minimum standard error, labor intensity, and cost of information processing has been studied. Implemented mechanisms for double threshold noise filtering, and detection of blurred image points for various practice conditions. A technique for gradient optimization of the micro-object identification functional, calculation of the weight coefficient of the image convolution kernel, and application of the Gauss, Laplace, Prewitt, Roberts, Sobel, and Canny operators is proposed. A software package for visualization, identification, recognition, and classification of images of micro-objects has been implemented, including a database of images, and subsystems for image pre-processing, filtering, identification, recognition, and classification. The operability of the software package was tested on the example of image processing of a large number of pollen grains and medical microorganisms.