Precision farming that takes advantage of the internet of things infrastructure now includes weed identification as a core component. Weeds now account for 45 percent of crop losses in farming because of competition with crops. This figure can be lowered with effective weed detecting technology. One of the most important areas of AI, known as deep learning (DL), is revolutionizing weed discovery for site-specific weed management (SSWM). In the past half a decade, DL methods have been used with both ground- and air-based technology for weed documentation in still images and in real time. According to the latest findings in DL-based weed detection, developing methods that aid precision weeding technologies in making informed decisions is a priority. Over the past five years, deep learning algorithms have been successfully incorporated into both ground-based and aerial-based systems for the purpose of weed identification in both still picture and real-time scenarios.
TopIntroduction
Agriculture is a vital aspect of the international economy since it generates substantial income and serves as a critical link in the distribution of food throughout the world. Because of this, agriculture has world's leading economic sectors. The agricultural industry has a major impact on the global economy (Razfar, et al., 2022). The number of linked devices is constantly increasing, and the IoT is rapidly becoming an integral part of our daily lives. Now that digital farming has emerged, farmers and other organisations have access to a streamlined and efficient method for real-time plant monitoring that combines creative tactics with state-of-the-art technology (Osorio, et al., 2020). Smart agriculture aims to reduce resource waste while raising crop yields via the application of technological advancements that augment conventional farming methods. This means that many parts of the production process can benefit from the use of technology on intelligent farms (Islam, et al., 2021). Invasive species are one of the most significant biotic constraints on crop construction. New approaches are being developed as part of an attempt to raise the total amount of soybeans collected by addressing the problem of weeds reducing yields' quality and quantity. The correctness or even the improvement of agro vision systems might be negatively impacted by the poor visibility that happens commonly during hazy or weather schemes (Asad, and Bais, 2020). During some weather patterns, visibility tends to be low.
Weeds in agriculture are defined as reducing crop yields and increasing production costs (Hasan et al., 2021). Weeds aren't usually crop-specific; however, they might share characteristics with annual, biennial, or perennial crops. It is easier to control annual weeds than perennial ones (Veeranampalayam et al., 2020). Plants that germinate from seed, mature for a solo growing season, and then perish are known as annual weeds. One of the most pressing problems that farmers confront today is preventing the spread of weeds (Haq., 2022). It is not uncommon for annual and perennial weeds endemic to Saudi Arabia, adjacent areas. Early detection and control, particularly prior to flowering, is the most effective method for eradicating weeds (Wu et al.,2021). Typically, people use their own eyes to look for weeds. To properly investigate and exert effort to reduce invasive weeds, it is necessary to monitor their spread and activity in near real-time (Shanmugam, et al., 2020).
As a result of its usefulness as a vegan protein source, soybean farming is flourishing. The success of such crops relies on the careful elimination of weeds. To be successful, SSWM must employ a range of weed management techniques, each of which must be tailored to the specific context, population, and density of the weeds in question. The machine's “brain,” a vision-based image processing system (Subeesh, et al., 2022), controls the rate at which herbicide is sprayed, allowing for such practises to be carried out with variable rate technology (VRT). One may find mapbased apps and sensorbased applications for VRT right now. A frequent method is map-based, in which a map of a region is created using georeferenced soil or plant samples. Collecting soil samples by hand for subsequent investigation is labor-intensive, time-consuming, and costly. However, sensor-based mapping can gather and interpret data in real time, making the process far more efficient. When using mobile ground or aerial technology in the field, all processing is done in real time through the use of machine learning or deep learning algorithms (Peng et al., 2022).