FinerForecasts - Biologically Driven Soft-Fruit Resource Optimisation, Labour & Yield Forecasts at Plant Granularity.

FinerForecast是由FruitCast牵头的一个合作项目，与英国肯特郡的软果种植者林肯大学和钱伯斯大学合作。该项目将使用基于观测的系统提高软果作物预测的准确性。作物预报对生鲜农产品商业运营至关重要，因为它使农民能够匹配供需、设定远期价格、减少浪费、计划劳动力和优化资源。现有的预测工具从种植者的直觉到机器学习支持的模型，都是出了名的不可靠，导致收入损失。FinerForecast将开发一个植物测绘系统，实现植物水平的预测，其中包括植物之间的变异性，提高整体预测精度，并使种植者能够在产量受到影响之前优化资源和解决问题点。这是对FruitCast目前基于观测的预测方法的扩展，该方法包括拍摄作物的视频，并在其中定位水果和花朵，以预测年龄、重量和收获日期。该项目旨在实现三个目标：(I)通过生物调控的产量预测模型，提前3周为整个种植地点提供可靠的15%误差范围内的产量预测；(Ii)为资源分配的优化策略生成预测产量及其变异性的植物水平、农学相关地图；以及(Iii)开发能够以每株为单位的分辨率在多个地点扩展开发的预测系统的数字架构，为英国市场做好准备。FinerForecast将产生更准确的产量预测，为当前的草莓市场带来高达5600万GB的收益(参见结果和市场路线)，并为种植者产生管理作物的新工具。该项目还旨在减少废物(目前为18kt)和这些废物中嵌入的二氧化碳排放(46kt CO2e)、水和杀虫剂，减少对环境的影响，提高生产率，并确保新鲜农产品部门更具弹性和可持续发展。FinerForecast利用FruitCast的能力，从视频中快速且廉价地测量作物状态，使商业规模的植物水平预测成为可能，从而提高预测精度。林肯大学在农业机器人和人工智能方面的专业知识将为该项目的成功做出贡献，钱伯斯的支持将使数据收集和系统部署成为可能。FinerForecast将通过提供准确和细粒度的产量预测，减少浪费和环境影响，优化资源，并提高新鲜农产品行业的弹性和可持续性，从而改变软果行业。

FinerForecasts is a collaborative project led by FruitCast, partnered with the University of Lincoln and Chambers, a soft fruit grower in Kent, UK. The project will improve the accuracy of soft fruit crop forecasting using observation-based systems. Crop forecasting is essential for fresh produce business operations as it enables farmers to match supply and demand, set forward prices, reduce waste, plan labour, and optimise resources. Existing forecast tools range from grower intuition to models powered by machine learning, and are notoriously unreliable, leading to loss of income. FinerForecasts will develop a plant mapping system, enabling plant-level forecasting, which incorporates variability between plants, improves the overall forecast accuracy, and enables growers to optimise resources and address problem spots before yield is impacted. This extends FruitCast's current observational-based forecasting approach involves taking videos of crops and locating fruit and flowers within them to predict the age, weight, and harvest date. The project aims to achieve three objectives: (i) provide reliable yield forecasts within 15% error 3 weeks ahead for entire grower sites from a biologically regulated yield forecasting model, (ii) generate plant-level, agronomically relevant maps of forecasted yield and its variability for optimisation strategies for resource allocation, and (iii) develop a digital architecture capable of scaling the developed forecasting system across multiple sites at a per-plant resolution, ready for the UK market. FinerForecasts will produce more accurate yield forecasts, contributing up to £56m(see\_outcomes\_and\_route\_to\_market) of benefits to the current strawberry market, and the generation of new tools for growers to manage crops. The project also aims to reduce waste (currently 18KT) and the CO2 emissions (46KT CO2e), water, and pesticides embedded in that waste, reducing environmental impact, driving productivity, and securing a more resilient and sustainable fresh produce sector. FinerForecasts leverages FruitCast's ability to quickly and cheaply measure crop state from videos to make plant-level forecasts possible at commercial scales, increasing forecast accuracy. The University of Lincoln's expertise in agri-robotics and AI will contribute to the project's success, and Chambers' support will enable data collection and deployment of systems. FinerForecasts will transform the soft fruit industry by providing accurate and granular yield forecasts, reducing waste and environmental impact, optimising resources, and improving the resilience and sustainability of the fresh produce sector.