A smart spraying system in agriculture is a targeted spraying system with efficient application of chemical and low cost for the environment. A smart sprayer generally includes a targeted detection system and spraying system, in which the targeted sensor is the foundation of the precision spraying management. The detection system of a smart sprayer is used to collect information in target areas and make spraying decisions. Varieties of sensing techniques are applied, such as Machine vision, spectral reflectance, remote sensing and so on. According to the detecting results of characteristics detection, species classification, disease symptom identification and damage severity evaluation, the spraying system controls the sprayer operation. The review of application of detection techniques, challenges and limitations are summarized, the developing trend is concluded based on the analysis.
Over-spraying herbicides puts pressing challenges on the agriculture industry. Farmers spend $25 billion per year buying 3 billion pounds of herbicides. But a huge volume of these chemicals never reaches weeds. Rather, it lands on soil or healthy plants or is carried away with rainwater. Applying conventional spraying technology, farmers lose money on herbicides that are sprayed in vain. Furthermore, these chemicals contaminate the soil, harming the environment; meanwhile, the weeds themselves develop herbicide resistance.
The reason for this inefficiency lies in the poor precision of broadcast sprayers. AgriTech providers claim that smart spraying solutions can lower the costs of herbicides by 90% due to selective application on weeds only. Some of the most prominent technologies enabling farmers to advance from broadcast to smart spraying are GPS guidance, machine learning algorithms, and computer vision for weed recognition.
Facebook is getting into farm equipment by training their AI to spot the weeds among the crops.
Farmers are always striving for bigger yields, but weeds get in the way. Uncontrolled weeds cause about about $43 billion in annual losses to soybean and corn farmers. While herbicides can do wonders at getting rid of weeds, they are costly, challenging to work with, and can kill valuable crops.
Facebook is teaming up with Pytorch and Blue River Technologies, a subsidiary of John Deere, to create farm equipment with artificial intelligence that will probe through fields identifying and killing weeds but leaving crops unharmed — bringing down the farmer’s cost in herbicides. This “see and spray” machine essentially uses the facial recognition technology Facebook is known for.
The AI farm equipment integrates machine learning and computer vision. It is built on PyTorch, an open-source, machine-learning platform created by Facebook’s AI research group. Since precision spraying has been shown to minimize herbicide resistance — something that will drive up a farmer’s annual herbicide costs — the team at Pytorch trained the AI to precisely spray only the weeds, leaving the crops intact. They did this by showing the program pictures of weeds until it could identify them among the plants.
The machine maps the field by pulling along cameras to examine the area frame by frame. Then, robots reference the map to accurately spray only the weeds.