Robotic High-Throughput Phenotyping (HTP) technology has been a powerful tool for selecting high-quality crop varieties among large quantities of traits. Due to the advantages of multi-view observation and high accuracy, ground HTP robots have been widely studied in recent years. In this paper, we study an ultra-narrow wheeled robot equipped with RGB-D cameras for inter-row maize HTP. The challenges of the narrow operating space, intensive light changes, and messy cross-leaf interference in rows of maize crops are considered. An in situ and inter-row stem diameter measurement method for HTP robots is proposed. To this end, we first introduce the stem diameter measurement pipeline, in which a convolutional neural network is employed to detect stems, and the point cloud is analyzed to estimate the stem diameters. Second, we present a clustering strategy based on DBSCAN for extracting stem point clouds under the condition that the stem is shaded by dense leaves. Third, we present a point cloud filling strategy to fill the stem region with missing depth values due to the occlusion by other organs. Finally, we employ convex hull and plane projection of the point cloud to estimate the stem diameters. The results show that the R² and RMSE of stem diameter measurement are up to 0.72 and 2.95 mm, demonstrating its effectiveness.

Weed classification is a fundamental perception task for agricultural robots and an essential enabler of precision and sustainable farming. Existing solutions often rely on high-power edge computing platforms, which limit long-term autonomous operation in Internet of Things (IoT) environments. Meanwhile, the computational complexity of high-accuracy deep learning (DL) models hinders their deployment on resource-constrained microcontrollers (MCUs), a critical component of IoT edge nodes. To address these challenges, this article proposes a tiny machine learning (TinyML)-enabled energy-efficient IoT framework for on-device weed classification, integrating a novel 3-D alignment knowledge distillation (TDA-KD) strategy with a lightweight multilayer dilated-convolution student network. The framework enhances knowledge transfer by jointly aligning: 1) individual predictions; 2) inter-sample correlations; and 3) class semantics, further strengthened through a multitemperature calibration mechanism. Experimental results on the DeepWeeds and 4-Weeds datasets demonstrate that the proposed student model achieves over 95% classification accuracy with only 240 K parameters and 87.51 MFLOPs. The model is successfully deployed on an OpenMV H7 Plus board with an STM32H7 MCU, requiring just 105.68 KB Flash memory and achieving an inference time of 378.3 ms with 510.7 mJ energy consumption per sample. A runtime analysis on the Vitirover horticultural robot shows that, compared with a Jetson Nano-based implementation, the proposed IoT pipeline extends operational time by approximately 30.5%. These results highlight the feasibility of deploying high-accuracy weed classification directly on ultralow-power IoT devices, thereby significantly enhancing the autonomy, energy efficiency, and scalability of agricultural robots.

With the advantages of high measurement accuracy and wide detection range, LiDARs have been widely used in information perception research to develop agricultural robots. However, the internal configuration of the laser transmitter layout changes with increasing sensor working duration, which makes it difficult to obtain accurate measurement with calibration files based on factory settings. To solve this problem, we investigate the intrinsic calibration of multi-beam laser sensors. Specifically, we calibrate the five intrinsic parameters of LiDAR with a nonlinear optimization strategy based on static planar models, which include measured distance, rotation angle, pitch angle, horizontal distance, and vertical distance. Firstly, we establish a mathematical model based on the physical structure of LiDAR. Secondly, we calibrate the internal parameters according to the mathematical model and evaluate the measurement accuracy after calibration. Here, we illustrate the parameter calibration with three steps: planar model estimation, objective function construction, and nonlinear optimization. We also introduce the ranging accuracy evaluation metrics, including the standard deviation of the distance from the laser scanning points to the planar models and the 3σ criterion. Finally, the experimental results show that the ranging error of calibrated sensors can be maintained within 3 cm, which verifies the effectiveness of the laser intrinsic calibration.

With the continuous growth of global population and increasing pressure on food security, the transformation toward precise and intelligent agricultural production has become an inevitable trend. In this context, accurate identification of field weeds is crucial for improving crop yields and reducing agricultural inputs. However, agricultural Internet of Things (IoT) edge devices are generally subject to strict constraints in terms of power consumption, storage, and real-time performance. Existing lightweight convolutional neural networks often struggle to simultaneously achieve high accuracy and low resource consumption for fine-grained weed identification tasks. To address this challenge, this paper proposes a hardware aware lightweight convolutional neural network named Weed-Aware Ghost Enhanced Network (WAGENet) for microcontroller deployment. The network synergistically integrates Ghost low-cost feature generation, Mobile Inverted Bottleneck Convolution (MBConv) for deep semantic extraction, Squeeze and Excitation (SE) and Coordinate Attention (CA) dual attention mechanisms for channel space joint calibration, and Atrous Spatial Pyramid Pooling (ASPP) for multi-scale context fusion. It constructs a progressive feature abstraction system from shallow textures to high-level semantics. On the public DeepWeeds dataset, WAGENet achieves 95.71% classification accuracy and 93.80% F1 score with only 0.163 M parameters and 2.43 × 10⁸ multiply accumulate operations (MACC), attaining a parameter efficiency of 587.19%/M and significantly outperforming existing mainstream lightweight models. The model has been successfully deployed on the STM32H7B3I microcontroller development board, achieving a single inference latency of 94.63 ms, an internal Flash footprint of only 686.95 KiB, and a single inference energy consumption of 41.45 mJ. Experimental results demonstrate that WAGENet achieves a trade off among accuracy, latency, and energy consumption under strict resource constraints, providing a reproducible microcontroller deployment paradigm for battery powered field robots, drones, and other agricultural IoT edge devices.

Underwater pipeline edge detection using underwater imaging sensors is the foundation for subsequent tasks such as pipeline positioning, tracking, defect detection and maintenance. However, this detection is often hindered by severe occlusion and extreme lighting conditions. Inspired by the human visual reasoning, the detection effect of pipeline edges under challenging scenarios primarily relies on contextual information. Therefore, we propose a context-enhanced deep learning network based on keypoint representation for underwater pipeline edge detection (CEPDNet). The network can effectively capture contextual information by fusing a criss-cross attention mechanism (CCAM) that is highly sensitive to the criss-cross distribution of underwater pipelines. Additionally, a slope output branch (SOB) is added to the CEPDNet to correlate each keypoint with its adjacent points for local information supplement. To fill the gap of diversified, multi-scene and multi-view underwater pipeline edge detection dataset, an annotated dataset (UPEdge) comprising real, generated, and simulated data has been constructed. Notably, an automatic annotation method in simulated environment is proposed. Experimental results show that our model achieves a detection accuracy of 91.32% on the self-constructed dataset. Moreover, our model demonstrates high deployment friendliness, with a detection speed of 4.9 ms per image post-deployment. In generalization tests, CEPDNet also achieves competitive performance on the TuSimple dataset.