List of Articles Najma Soltani

• Open Access Article
  • Abstract Page
  • Full-Text
  
  1 - Deep learning approach to help Chenopodiaceae biodiversity protection to prevent soil erosion (case study: Yazd province, Iran)
  Ahmad Heidary-Sharifabad Najma Soltani
  10.30495/jonass.2022.1953114.1035
  Background and objective:Chenopodiaceae species are important vegetation around the world, especially in the desert and semi-desert areas. Preserving the biodiversity of Chenopodiaceae species is crucial to preventing soil erosion. In addition, most of them are of ecolo More

  Background and objective:Chenopodiaceae species are important vegetation around the world, especially in the desert and semi-desert areas. Preserving the biodiversity of Chenopodiaceae species is crucial to preventing soil erosion. In addition, most of them are of ecological and economic importance and also play an important role in biodiversity around the world. Conservation of this biodiversity is vital to the survival and sustainability of the ecosystem. To protect plant biodiversity, it is essential to know the plant species in their natural habitats. Therefore, automatic identification of plant species in their habitat helps to analyze the species and thus take care of their biodiversity. Computer vision approaches can be used to automatically identify and classify plant species. Modern approaches use deep learning in computer vision.Materials and methods:In this study, the ACHENY data set that consists of 27030 images of 30 species of Chenopodiaceae are used. Firstly, using the SuperPixel method, larger size images (448×448) than existing ACHENY dataset images size (224×224) are created. Secondly, based on the newly created dataset we introduce a proper deep learning model to identify Chenopodiaceae species.Results and conclusion:The results of the evaluation confirm the improvement of the classification accuracy of ACHENY species by the proposed model compared to the previously presented models. The results of the experiments indicate a superiority of about 3% accuracy of the proposed method and all evaluation parameters of the research have increased to a reasonable extent. Manuscript profile