Mustansiriyah journal of pure and applied sciences

Plants are the main source of the production of herbal volatile oils, followed by animal sources of amber, which is extracted from the amber whale, and musk, which is extracted from deer. The plantae kingdom includes more than 2000 plants containing volatile aromatic oils, which represent about sixty plant families, the most important of which are: Rosaceae, Asteraceae, Myrtaceae, Lamiaceae, Apiaceae, Rutaceae, Leuraceae, Pinaceae and Geraniaceae.Aromatic oils, which are organic compounds that are responsible for giving the plant a distinctive smell and they are easily separated from the plant parts containing them by different extraction methods. They are called volatile oils because they are characterized by evaporation and volatilization without decomposition at room temperature and they are also called essential oils for the human need for them in his food and the inability of the enzymatic cellular system in the human body to synthesize them or ethereal oil because it is soluble in alcoholic compounds such as ether, and it does not contain in its molecules fatty acids or glycerin, so it does not rancid when exposed to direct air or light, unlike fixed oils that do not volatilize, but decompose when exposed to high temperatures. Volatile oils are found in the tissues of various plant parts such as leaves, flowers, roots and stems.

The significance of content-based annotation in object image recognition is critical for monitoring applications. With the rapid advancements in remote sensing technologies, the abundance of data generated by advanced objects, such as satellite imagery, presents valuable resources for a multitude of real-world applications, from urban planning to ecological monitoring [1]. The vastness of this data repository presents both opportunities and challenges. In this study, we focus on label image categorization, aiming to assign remote sensing images to a set of predefined object-level tags. Automated annotation not only expedites image retrieval but also provides semantic concepts for diverse applications. Label scene recognition is common in computer vision applications[2]. In real-world scenarios, images can encapsulate a variety of correlated tags and convey objects within a region of interest. Although deep learning methods have transformed image processing and computer vision, their application in object image recognition remains nascent, particularly because of the scarcity of labeled datasets[3]. This study addresses this gap by aiming to enhance label recognition. However, these challenges persist in the future. Extracting object-level features for label tasks remains complex, given the variation in label visibility across images. Advances in label dependencies, such as sequence- and graph-based approaches, have emerged to enhance recognition capabilities[4]. Reliable labels in a sequence or graph can serve as predictive information for the other labels. Notably, the distribution of concepts across modalities (image, graph, and text) presents opportunities for cross-modal feature enhancement. Object detection is intended to locate and classify an object within a specific picture or frame in video format. This objective is accomplished by merging several complicated operations, such as object localization and classification. Different methodologies have been used to address these problems. Nevertheless, progress in Deep Learning, including neural network architectures, has opened up a new epoch of Object Detection, where convolutional neural networks (CNNs) have become an industry’s standard for image classification tasks; meanwhile, Recurrent Neural Networks (RNN The current paper presents a novel CNN-LSTM hybrid architecture, which capitalizes on the complementary strengths of both paradigms, and offers an all-inclusive response to image classification problems.