Key Technology Research On Strawberry Picking Robot In Greenhouse Environment Under Elevated Cultivation

In the past decade,China has become the world’s largest producer and consumer of strawberries,with a continuous increase in planting area and consumption.However,with the rapid growth of labor cost,the existing manual work-based planting and management mode has become one of the important factors limiting the development of strawberry industry.It is important to reduce the labor requirement of strawberry production and improve the automation and intelligence of the production process to promote the sustainable development of strawberry industry.The automation of picking operation can effectively reduce the manpower demand in the strawberry production process,which is the central part of the strawberry production process.The basis of automatic fruit picking is to be able to obtain the spatial position of strawberry targets stably.There are interference factors such as light intensity variation,fruit overlapping and branch shading in natural environment,so strawberry target identification,localization and spatial pose estimation in natural environment are the key technologies of strawberry automatic picking equipment.In this paper,the following research is conducted for the identification and localization method of strawberry fruits cultivated at high price in greenhouse environment.1.In the target recognition and localization system,the strawberry target recognition and fine segmentation technology based on image target prepositioning and combined with convolutional neural network and U-net network is proposed to realize the high precision localization of strawberry targets and the accurate segmentation of strawberry target point cloud.2.In the judgment of strawberry ripeness and occlusion,the judgment of strawberry target ripeness and whether it is occluded or not is completed based on the convex packet feature of strawberry target region image and combined with support vector classifier.3.Based on the geometric characteristics of the target point cloud and the strawberry target surface point cloud obtained from the strawberry target area fine segmentation system,we complete the estimation of position and picking points for intact and obscured strawberries.