Agriculture, practised for thousands of years, produces food for human consumption and products for industrial use, by growing crops in fields or indoor controlled environments. Crop plants face nutrient deficiencies, weed competition, and pest and disease attacks. These are typically managed by applying liquid treatments of fertilizers and pesticides. To move away from treating entire fields and large areas, we must pursue the ultimate goal of Precision Agriculture by treating individual plants or leaves. Two technologies are necessary for such 'ultra-precision. ' First, a vision system must detect problems within the crop, e.g. leaf parts, disease lesions, insect clusters, or weeds. Second, small amounts of liquid must be applied to target areas without touching the surrounding crop or environment. AgriSynth has developed a vision system that can identify objects in crops at a millimetre scale. Most farms still apply liquid chemical sprays to entire fields or sections of fields. Only the most advanced machinery can target areas as small as one square meter. With agricultural robots, this precision can be increased to 200-300 square centimetres. However, for targets millimetres in size, this level of precision is insufficient, leading to the treatment of unaffected areas, wasted chemicals, and unnecessary environmental impact. Imagine placing a football in a wheat crop (similar to a domestic lawn) so it touches a weed but not the crop, then a tennis ball, then a golf ball. In most situations, only when we get to the size of a pea can we touch a weed, not the crop. This is the treatment resolution we need to apply a liquid to a 5mm-diameter target area at a range of approximately 200mm for most Use Cases. We can't achieve this in agriculture, but this project aims to change that. The concept will be like how inkjet printers revolutionized printing on paper. Developing an inkjet-like applicator for crops will achieve ultra-precision. If designed modularly, the system could be mounted on existing machinery, robots, and even handheld devices providing benefits in agricultural, horticultural, domestic and a range of other markets. This innovation aims to enable extremely low-dose applications, eliminating soil contact and water contamination. By targeting specific weeds, we could use liquid treatments not traditionally approved for crops, by avoiding touching the crop plants. This project could revolutionize crop productivity, enabling farmers to manage crops at the plant and leaf level, significantly improving yields and almost eliminating environmental impact.

Agricultural robots require effectively trained AI systems to carry out functions effectively. The agricultural sector is one of the most difficult in which to train AI systems to interpret agricultural scenes due to multiple layers of complexity: * **Plant/weed species:** huge species variances and multiple species both difficult to distinguish at early growth stages * **Occlusion:** In complex crop scenes many crop and weed plants overlap in a complex manner * **Physical changes:** Effects of pests/diseases, leaf/crop deformities and soil changes * **Different presentations:** Camera angles, scene lighting and backgrounds create variabilities and translucency effects * **Annotation:** Annotation of images at pixel level is almost impossible for humans to do accurately and at volume One of the most difficult and economically damaging problems for UK farmers is blackgrass, which threatens the viability of wheat crops. Blackgrass is difficult for an agricultural robot to detect/distinguish at the early stage which is required for effective elimination treatments as this requires a significant, varied, dataset which could take years to obtain. Such a robust AI solution does not exist today. During this project, we will develop an advanced synthetic image modelling engine capable of automatically producing high volumes of variable complex crop scene datasets at <10% of real-world costs. These can be used to effectively train AI solutions within days rather than months/years. This will enable agricultural robots to robustly detect blackgrass. Blackgrass is the first use case and further development within the project will enable other species to be classified in a wide range of conditions.