CTDA: An accurate and efficient cherry tomato detection algorithm in complex environments

Zhi Liang ¹ Caihong Zhang ² Zhonglong Lin ¹ Guoqiang Wang ² Xiaojuan Li ^1* Xiangjun Zou ^1*

• ¹ Xinjiang University, Urumqi, China
• ² Institute of Agricultural Mechanization, Xinjiang Academy of Agricultural Sciences, Urumqi, Xinjiang, China

In the natural harvesting conditions of cherry tomatoes, the robotic vision for harvesting faces challenges such as lighting, overlapping, and occlusion among various environmental factors. To ensure accuracy and efficiency in detecting cherry tomatoes in complex environments, the study proposes a precise, real-time, and robust target detection algorithm: the CTDA model, to support robotic harvesting operations in unstructured environments. The model, based on YOLOv8, introduces a lightweight downsampling method to restructure the backbone network, incorporating adaptive weights and receptive field spatial characteristics to ensure that low-dimensional small target features are not completely lost. By using softpool to replace maxpool in SPPF, a new SPPFS is constructed, achieving efficient feature utilization and richer multi-scale feature fusion. Additionally, by incorporating a dynamic head driven by the attention mechanism, the recognition precision of cherry tomatoes in complex scenarios is enhanced through more effective feature capture across different scales. According to experimental findings, CTDA has strong robustness and adaptability in complicated environments. The detection accuracy reaches 94.3%, with a recall rate and mAP@0.5 of 91.5% and 95.3% respectively, while mAP@0.5:0.95 is 76.5% and FPS is 154.1 frames per second. Compared to YOLOv8, its mAP@0.5 has increased by 2.9% while maintaining detection speeds, with a model size of 6.7M, deployable on edge devices for rapid detection, and providing technical support for automated harvesting robots for greenhouse cherry tomatoes.