A Thresholding Approach for Pollen Detection in Images Based on Simulated Annealing Algorithm

Abstract

Melissopalynology is a field that studies pollen grain origins to identify their species. It consists of studying either the chemical composition of each grain, or their shapes using microscopic images. This paper presents a system of pollen identification based on the microscopic images, it is divided into two parts, first part is the pollen detection using a thresholding method with simulated annealing algorithm. The second step is the pollen classification, in which we used deep convolutional neural network to extract features from the detected pollen grains and represent them in numerical vectors, therefore, we can use these vectors to classify them based on fully connected neural network, SVM or similarity calculation. The obtained results showed a high efficiency of the neural network in which it could recognize 98.07% of the pollen species compared not just to SVM and similarity methods, but also to works from literature.

Automatic Image Classification Using Neural Networks Increases Accuracy for Allergenic Pollen Monitoring

“…Deep learning models have shown similar accuracy rates to ours on larger and more varied pollen datasets as well, but these either focussed on the family level 28 or on insect-collected pollen for honey analysis 29,30 . Increasing the taxonomic resolution of pollen grains has been achieved by incorporating an extensively trained deep learning model with super-resolution microscopy on a case study of fossil pollen 31 .…”