Prolonged operation of cotton picker poses significant risks of work-related musculoskeletal disorders (WMSDs), primarily driven by non-ergonomic cab layouts that fail to accommodate the unique “left-hand steering, right-hand lever” operational mode. Traditional optimization methods, relying on general digital human models or isolated surface electromyography (sEMG) measurements, often lack the physiological fidelity and computational efficiency for high-dimensional, personalized design. To address this interdisciplinary challenge in agricultural engineering and ergonomics, this study proposes a novel AMS-RF-DBO framework that integrates high-fidelity biomechanical simulation with intelligent optimization. A driver–cabin biomechanical model was developed using the AnyBody Modeling System (AMS) and validated against sEMG data (ICC = 0.695). This model generated a dataset linking cab layout parameters to maximum muscle activation (MA). Using steering wheel and control lever coordinates (X, Y, Z) as inputs, a Random Forest (RF) regression model demonstrated strong performance (R² = 0.91). Optimization with the Dung Beetle Optimizer (DBO) algorithm yielded an optimal configuration: steering wheel (L1 = 434 mm, H1 = 738 mm, θ = 32°) and control lever (L2 = 357 mm, H2 = 782 mm, M = 411 mm), reducing upper-limb MA from 3.82% to 1.47% and peak muscle load by 61.5%. This study not only provides empirical support for ergonomic cab design in cotton pickers to reduce operator fatigue and health risks but also establishes a replicable technical paradigm for ergonomic optimization of other specialized agricultural machinery.

Automatic container-number recognition in land-port yards is challenged by clutter, illumination changes, scale variation, surface wear, and character distortion. This paper proposes a two-stage detection–cropping–recognition framework comprising an improved YOLO11 detector and a lightweight convolutional recurrent neural network (CRNN). The detector integrates C3k2 with channel-split multiscale aggregation (C3k2-CSMA), a hierarchical efficient feature fusion pyramid network (HEFFPN), selective boundary aggregation (SBA), and a shared enhanced detection head (SED-Head) with detail-enhanced convolution (DEConv) to improve small-target representation, feature fusion, and boundary localization. The recognizer combines MobileNetV3 with bidirectional long short-term memory (Bi-LSTM) and is trained using connectionist temporal classification (CTC) loss and L2 regularization. The detector achieved 97.3% mean average precision at an intersection-over-union threshold of 0.5 (mAP@0.5), 97.1% recall, 55 frames per second (FPS), and a model size of 12.2 MB. The recognizer achieved 97.6% exact-match accuracy, a model size of 6.2 MB, and an inference time of 8.9 ms/crop. The complete system achieved 96.8% end-to-end exact-match accuracy, a total model size of 18.4 MB, and a total inference time of 27.1 ms/image, demonstrating a practical accuracy–efficiency balance for the evaluated complex land-port scenes.