An Improved Method for Palladium Nanoparticles Detection on Carbon Materials in Scanning Electron Microscope Images

The subject of this paper is the urgent problem of detecting palladium nanoparticles on car-bon materials in grayscale scanning electron microscope images, the solution of which can be useful in many technological processes of the chemical industry. This is nontrivial problem and, currently, there is no method that would have all the necessary qualities for its solution. This work is based on our previously developed method for detecting nanoparticles based on exponential approximation, which showed the best accuracy for scanning electron microscope images in absence of essential background irregularities. However, it has two significant shortcomings: 1) high demands on computing resources and 2) in the presence of background irregularities it produces a large number of false positives. This paper proposes a method that preserves the basic assumptions and general concept and, accordingly, the main advantages of earlier proposed approach, however, it contains a number of significant improvements that allow eliminating the above-mentioned shortcomings. Experiments show that the proposed method reduces the number of false positives by 60-70% for SEM images with background irregularities while decreasing the nanoparticle detection time by approximately two to three times compared with the previously proposed approach.