Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Abstract. Strawberries are an important economic crop. Precise counting of strawberry fruits can assist growers in estimating yield and arranging product flow into the market in advance. In this study, we proposed a dynamic object counting method for strawberry fruits from videos captured using a cell phone. Firstly, an improved YOLOv5s was developed by integrating a vision transformer layer to detect strawberry fruits in each frame from the input video. Secondly, we then proposed a dynamic counting algorithm to count the total number of fruits in the video. The counting algorithm consists of two modules: StrongSORT tracking and DynCount. The StrongSORT framework, a state-of-the-art tracking algorithm based on Kalman filter and Hungarian matching, is used to track the identities of strawberry fruits in the video data, while the DynCount algorithm maps the tracking IDs to counting numbers, and displays the counting numbers along with bounding boxes. Experimental results with ten videos showed that our approach achieved an average counting accuracy of 92.05%. Our method provides accurate and timely strawberry counting information, which can alleviate the labor burden and provide important reference indicators for yield prediction and market planning.