Drives & Controls March 2023

34 n FOOD AND BEVERAGE March 2023 www.drivesncontrols.com Farmers’ interest in machine vision is mushrooming T he mushroom industry is a booming global business, which was worth more than $50bn in 2019. The biggest producers are China and the US. To create a more favourable environment for growing mushrooms, farmers are increasingly moving their farming operations from outdoors into large greenhouses, thus shifting environmental influences from external climates to more controllable internal microclimates. A major driver behind this move has been the changing climate which has made outdoor farming less productive than it used to be. Controlling a greenhouse climate to optimise mushroomgrowth presents several challenges for farmers. Frequent temperature changes are needed for achieve a successful harvest. For example, the greenhouse temperature needs to be reduced from 22°C to 16°C to stimulate fruiting. Moreover, relative humidity has to be maintained at between 85% and 90%, and carbon dioxide concentrations need to be adjusted to appropriate levels at each of the mushrooms’six growth stages. Traditionally, farmers have drawn on their personal experience and visual observations to estimate the relationship between mushroom growth and greenhouse microclimates, and relied on their expertise to determine the optimum temperature, humidity and other environmental factors. Greenhouses are usually equipped with environmental monitoring systems to obtain data on their microclimate, however, there is no sensor that can measure the growth of mushrooms directly to determine when adjustments need to be made. Recently, computer vision technologies have started to be applied to agriculture – notably, deep-learning neural networks that can recognise individual images among multiple objects, such as a single mushroom cap in a large field of mushrooms. Harnessing the power of neural networks, researchers at Meiho University in Taiwan have developed a machine vision algorithm that can analyse images of the entire fruiting period of individual mushrooms. By recording the size of individual mushroom caps continuously, data can be generated that is used to optimise the greenhouse climate controls, to calculate growth rates, and to act as harvest reminders. During the experiment, mushroom images were captured using SVS-Vistek 1.2-megapixel GigE Vision colour cameras with a resolution of 1296 × 964 pixels and a 30 frames-per- second capture rate. Besides their accurate imaging, these cameras were chosen because they are rugged enough to survive in a greenhouse climate which presents similar stresses to an outdoor installation, including large variations in humidity and heat. The cameras are IP67-protected – unusual for machine vision systems – making them dust- tight and waterproof, as well as being resistant to vibration because of their use of M12 connectors. The scientists installed the cameras 25cm above a mushroom cultivation bed with a light source at the top of the greenhouse. The cameras measured the growth of the mushrooms automatically during the fruiting period, calculating the diameters of the mushroom caps once an hour. This data was multiplied by the spatial resolution to obtain the actual size of the mushroom caps. The researchers conducted experiments and then compared their results with traditional imaging methods to validate their technology. They concluded that the new algorithm can successfully analyse images of the entire fruiting period of the mushrooms, making it attractive to farmers because it can make is easier for them to control the microclimates in their greenhouses. The researchers plan to expand their experiments to test the algorithm in large-scale commercial applications. n Researchers in Taiwan are using industrial machine vision cameras to monitor the growth of mushrooms in greenhouses, allowing farmers to tailor the environmental conditions to optimise the growth of the fungi. Mushrooms growing in the university laboratory. Image: Meiho University The machine vision cameras and software algorithms can identify individual mushrooms caps and track their growth even when they overlap with others Image: Meiho University