To fill this gap, in this paper, we proposed an innovative and effective indicator called distance-entropy, to assess data quality for the machine vision recognition tasks in agriculture. That is to say, there are limited research on image quality and information assessment aiming at AI-driven visual tasks. et al., 2021) and stereoscopic video ( Yang et al., 2019 Zhao et al., 2021). However, the current image quality assessment (IQA) research works are mainly focused on visual evaluation, including screen images ( Yang et al., 2020 Yang J. Therefore, the small amount of data must be built on the premise of high quality to be meaningful. The related research works are mainly meta-learning, model fine-tuning, and applications ( Karami et al., 2020 Nuthalapati and Tunga, 2021 Yang Y. But, most of the existing related studies in the literature are based on the randomly selected few data, without enough consideration of data information value. To explore the effect of data quality, a machine learning based on limited data, also called few-shot learning, has made some attempts and has emerged in many scenarios ( Li and Yang, 2020, 2021 Chao and Zhang, 2021 Li and Chao, 2021 Li et al., 2021 Nie et al., 2021). ![]() Therefore, the data quality analysis and screening are significant for practical applications. On the other hand, edge computing ( Yang et al., 2018 Huang et al., 2021) has relatively weak hardware resources and high training costs. The data quality should be treated seriously as this is gradually becoming a research focus, thus, helping to improve the training efficiency of models. For cloud computing ( Zuo et al., 2016 Zhu et al., 2017), the required communication bandwidth of data transmission is very high for real-time performance. Even in scenarios with big data, there is also an inevitable problem in the early stages of data collection.Ĭonsidering the execution of AI-based algorithms, they are mainly cloud computing and edge computing. ![]() However, the limited amount of data is the essential property of many real-world tasks. One main drawback is that deep models are unfriendly to a small amount of data and have a severe over-fitting problem. For example, based on the RGB image processing or hyperspectral image processing, there have been numerous typical studies and applications, including agricultural yield forecasting ( Khaki and Wang, 2019 Shahhosseini et al., 2020 Jarlan et al., 2021), crop pests and diseases identification ( Li and Chao, 2020 Li et al., 2020 Liu and Wang, 2020 Liang, 2021), agricultural robot and navigation ( Wen et al., 2020 Zhang et al., 2020 Emmi et al., 2021), counting of plant fruits ( Lin and Guo, 2020 Fu et al., 2021), etc.ĭeep learning is the primary implementation of intelligent applications by combining ICT and agriculture, but the shortcomings are also apparent. However, more commonly used data sources in artificial intelligence (AI)-driven applications are images or videos. As for the intelligent plant protection, the related data have various sources, such as the sensing of soil ( Yin et al., 2021), light intensity ( Yu et al., 2021), water stress ( Mundim and Pringle, 2018 Ihuoma and Madramootoo, 2019), mixture of water and fertilizer ( Jia et al., 2019), temperature and humidity ( Mekala and Viswanathan, 2019), etc. Smart agriculture is established based on the digital process, combining the data in the agricultural field and the Information and Communications Technology (ICT) ( Friha et al., 2021 Sun et al., 2021). In general, this study is a relatively cutting-edge work in smart agriculture, which calls for attention to the quality assessment of the data information and provides some inspiration for the subsequent research on data mining, as well as for the dataset optimization for practical applications. Both the numerical and the visual results demonstrate the effectiveness and stability of the proposed distance-entropy method. Many comparative experiments, considering the mapping feature dimensions and base data sizes, were conducted to testify the validity and robustness of this indicator. In this work, taking the crop pest recognition task as an example, we proposed an effective indicator of distance-entropy to distinguish the good and bad data from the perspective of information. However, the current image quality assessment research focuses on visual quality, while ignoring the crucial information aspect. A high-quality data improves the efficiency of intelligent algorithms and helps reduce the costs of data collection and transmission. Smart agriculture is inseparable from data gathering, analysis, and utilization. 2School of Electrical and Information Engineering, Tianjin University, Tianjin, China.1College of Mechanical and Electrical Engineering, Shihezi University, Xinjiang, China.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |