Institut für Phytomedizin

Permanent URI for this collectionhttps://hohpublica.uni-hohenheim.de/handle/123456789/14

Browse

Now showing 1 - 2 of 2

Assessing the capability of YOLO- and transformer-based object detectors for real-time weed detection
(2025) Allmendinger, Alicia; Saltık, Ahmet Oğuz; Peteinatos, Gerassimos G.; Stein, Anthony; Gerhards, Roland
Spot spraying represents an efficient and sustainable method for reducing herbicide use in agriculture. Reliable differentiation between crops and weeds, including species-level classification, is essential for real-time application. This study compares state-of-the-art object detection models-YOLOv8, YOLOv9, YOLOv10, and RT-DETR-using 5611 images from 16 plant species. Two datasets were created, dataset 1 with training all 16 species individually and dataset 2 with grouping weeds into monocotyledonous weeds, dicotyledonous weeds, and three chosen crops. Results indicate that all models perform similarly, but YOLOv9s and YOLOv9e, exhibit strong recall (66.58 % and 72.36 %) and mAP50 (73.52 % and 79.86 %), and mAP50-95 (43.82 % and 47.00 %) in dataset 2. RT-DETR-l, excels in precision reaching 82.44 % (dataset 1) and 81.46 % (dataset 2) making it ideal for minimizing false positives. In dataset 2, YOLOv9c attains a precision of 84.76% for dicots and 78.22% recall for Zea mays L.. Inference times highlight smaller YOLO models (YOLOv8n, YOLOv9t, and YOLOv10n) as the fastest, reaching 7.64 ms (dataset 1) on an NVIDIA GeForce RTX 4090 GPU, with CPU inference times increasing significantly. These findings emphasize the trade-off between model size, accuracy, and hardware suitability for real-time agricultural applications.
Sensor-guided mechanical weed control in transplanted lettuce and cabbage
(2025) Gerhards, Roland; Spaeth, Michael; Heyn, Alexandra; Saile, Marcus
Effective weed control is extremely important in vegetable production because weeds affect yield and quality of vegetable crops. Usually, only combinations of preventive and direct weed control methods can sufficiently suppress weeds. Therefore, costs for weeding are much higher in vegetables compared to most arable crops. Due to restrictions for herbicide use in vegetables, alternative and efficient direct weeding methods are urgently needed. Six field experiments with transplanted cabbage and lettuce were conducted in Southwestern Germany to quantify the weed control efficacy (WCE) and crop response of sensor-guided mechanical weed control methods with different degrees of automation. A camera-guided inter-row hoe with automatic side-shift control alone and combined with intra-row finger weeders and a camera-guided intra-row hoeing (robot) were compared to standard mechanical weeding, a broadcast herbicide treatment and an untreated control. Weed densities prior to treatment averaged 58 plants m −2 in cabbage and 18 plants m −2 in lettuce. Chenopodium album, Amaranthus retroflexus, Thlaspi arvense, Solanum nigrum and Digitaria sanguinalis were the dominating species. Until harvest, 80% weed coverage was measured in the untreated plots of cabbage and 28% in lettuce, which caused 56% yield loss in cabbage and 28% yield loss in lettuce. The highest WCE was achieved with the robot (87% inter-row and 84% intra-row) The broadcast herbicide treatment achieved 84% WCE for both inter-row and intra-row areas. Conventional inter-row hoeing had the lowest WCE of 73% inter-row and 35% intra-row. Camera-guided inter-row hoeing increased inter-row WCE to 80% and intra-row WCE to 56%. Finger weeding increased intra-row WCE in lettuce and cabbage to 54% with conventional hoeing and 62% with camera-guided hoeing. Camera-guidance reduced crop plant losses by 50% (from 9.1 to 4.5%) and increased crop yield by 13% compared to conventional hoeing. This study highlights the benefits of camera-guidance, AI-based weed detection and robotic weeding in transplanted vegetable crops.

Browse

Browsing Institut für Phytomedizin by Sustainable Development Goals "9"