Search

Copyright of Journal of Intelligent Agriculture Mechanization is the property of Nanjing Institute of Agricultural Mechanization and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Due to the complex shape of the tea tree canopy and the large undulation of a tea garden terrain, the quality of fresh tea leaves harvested by existing tea harvesting machines is poor. This study proposed a tea canopy surface profiling method based on 2D LiDAR perception and investigated the extraction and fitting methods of canopy point clouds. Meanwhile, a tea profiling harvester prototype was developed and field tests were conducted. The tea profiling harvesting device adopted a scheme of sectional arrangement of multiple groups of profiling tea harvesting units, and each unit sensed the height information of its own bottom canopy area through 2D LiDAR. A cross-platform communication network was established, enabling point cloud fitting of tea plant surfaces and accurate estimation of cutter profiling height through the RANSAC algorithm. Additionally, a sensing control system with multiple execution units was developed using rapid control prototype technology. The results of field tests showed that the bud leaf integrity rate was 84.64%, the impurity rate was 5.94%, the missing collection rate was 0.30%, and the missing harvesting rate was 0.68%. Furthermore, 89.57% of the harvested tea could be processed into commercial tea, with 88.34% consisting of young tea shoots with one bud and three leaves or fewer. All of these results demonstrated that the proposed device effectively meets the technical standards for machine-harvested tea and the requirements of standard tea processing techniques. Moreover, compared to other commercial tea harvesters, the proposed tea profiling harvesting device demonstrated improved performance in harvesting fresh tea leaves. [ABSTRACT FROM AUTHOR]

Cutting-edge detection is a critical step in mechanized rice harvesting. Through visual cutting-edge detection, an algorithm can sense in real-time whether the rice harvesting process is along the cutting-edge, reducing loss and improving the efficiency of mechanized harvest. Although convolutional neural network-based models, which have strong local feature acquisition ability, have been widely used in rice production, these models involve large receptive fields only in the deep network. Besides, a self-attention-based Transformer can effectively provide global features to complement the disadvantages of CNNs. Hence, to quickly and accurately complete the task of cutting-edge detection in a complex rice harvesting environment, this article develops a Transformer Convolution Network (TCNet). This cutting-edge detection algorithm combines the Transformer with a CNN. Specifically, the Transformer realizes a patch embedding through a 3 × 3 convolution, and the output is employed as the input of the Transformer module. Additionally, the multi-head attention in the Transformer module undergoes dimensionality reduction to reduce overall network computation. In the Feed-forward network, a 7 × 7 convolution operation is used to realize the position-coding of different patches. Moreover, CNN uses depth-separable convolutions to extract local features from the images. The global features extracted by the Transformer and the local features extracted by the CNN are integrated into the fusion module. The test results demonstrated that TCNet could segment 97.88% of the Intersection over Union and 98.95% of the Accuracy in the unharvested region, and the number of parameters is only 10.796M. Cutting-edge detection is better than common lightweight backbone networks, achieving the detection effect of deep convolutional networks (ResNet-50) with fewer parameters. The proposed TCNet shows the advantages of a Transformer combined with a CNN and provides real-time and reliable reference information for the subsequent operation of rice harvesting. [ABSTRACT FROM AUTHOR]

The yield and quality of rice are closely related to field management. The automatic identification of field abnormalities, such as diseases and pests, based on computer vision currently mainly relies on high spatial resolution (HSR) images obtained through manual field inspection. In order to achieve automatic and efficient acquisition of HSR images, based on the capability of high-throughput field inspection of UAV remote sensing and combining the advantages of high-flying efficiency and low-flying resolution, this paper proposes a method of "far-view and close-look" autonomous field inspection by unmanned aerial vehicle (UAV) to acquire HSR images of abnormal areas in the rice canopy. First, the UAV equipped with a multispectral camera flies high to scan the whole field efficiently and obtain multispectral images. Secondly, abnormal areas (namely areas with poor growth) are identified from the multispectral images, and then the geographical locations of identified areas are positioned with a single-image method instead of the most used method of reconstruction, sacrificing part of positioning accuracy for efficiency. Finally, the optimal path for traversing abnormal areas is planned through the nearest-neighbor algorithm, and then the UAV equipped with a visible light camera flies low to capture HSR images of abnormal areas along the planned path, thereby acquiring the "close-look" features of the rice canopy. The experimental results demonstrate that the proposed method can identify abnormal areas, including diseases and pests, lack of seedlings, lodging, etc. The average absolute error (AAE) of single-image positioning is 13.2 cm, which can meet the accuracy requirements of the application in this paper. Additionally, the efficiency is greatly improved compared to reconstruction positioning. The ground sampling distance (GSD) of the acquired HSR image can reach 0.027 cm/pixel, or even smaller, which can meet the resolution requirements of even leaf-scale deep-learning classification. The HSR image can provide high-quality data for subsequent automatic identification of field abnormalities such as diseases and pests, thereby offering technical support for the realization of the UAV-based automatic rice field inspection system. The proposed method can also provide references for the automatic field management of other crops, such as wheat. [ABSTRACT FROM AUTHOR]

The recognition of fresh tea leaf sprouts is one of the difficulties in the realization of the automated picking of fresh tea leaves. At present, the research on the detection of fresh tea leaf sprouts is based on a single variety of tea leaves for a specific period or specific place, which has no advantage for the spread, promotion, and application of the methods. To address this problem, an identification of multiple varieties of tea sprouts (IMVTS) model was proposed. First, images of three different varieties of tea (ZhongCha108 (ZC108), ZhongHuangYiHao (ZH), ZiJuan (ZJ)) were obtained, and the multiple varieties of tea (MVT) dataset for training and validating models was created. In addition, the detection effects of adding a convolutional block attention module (CBAM) or efficient channel attention (ECA) module to YOLO v7 were compared. In the detection of the MVT dataset, YOLO v7+ECA and YOLO v7+CBAM showed a higher mean average precision (mAP) than YOLO v7, with 98.82% and 98.80%, respectively. Notably, the IMVTS model had the highest AP for ZC108, ZH, and ZJ compared with the two other models, with 99.87%, 96.97%, and 99.64%, respectively. Therefore, the IMVTS model was proposed on the basic framework of the ECA and YOLO v7. To further illustrate the superiority of the model, this study also conducted a comparison test between the IMVTS model and the mainstream target detection models (YOLO v3, YOLO v5, FASTER-RCNN, and SSD) and the IMVTS model on the VOC dataset, and it is clear from the test results that the mAP of the IMVTS model is ahead of the remaining models. Concisely, the detection accuracy of the IMVTS model can meet the engineering requirements for the automatic harvesting of autumn fresh famous tea leaves, which provides a basis for the future design of detection networks for other varieties of autumn tea sprouts. [ABSTRACT FROM AUTHOR]

Tillage is one of the most important management practices in rice (Oryza sativa L.) production. However, reports comparing fragrant rice performance under plow tillage vs. rotary tillage are scarce. Here we conducted a five‐season field experiment with a fragrant rice cultivar, Meixiangzhan‐2, to study the effects of continuous plow tillage on fragrant rice yield, quality, and grain concentration of 2‐acetyl‐1‐pyrroline (2‐AP), compared with conventional rotary tillage. The results showed that after three cropping seasons of continuous plowing, compared with conventional rotary tillage, plow tillage not only significantly increased grain yield and affected panicle number per area by 2.98–11.43% and 4.45–18.81%, respectively, but also promoted grain 2‐AP concentration and protein content by 17.27–31.46% and 3.94–7.10%, respectively. Moreover, plow tillage decreased grain amylose content by 1.83–3.52%. Compared with conventional rotary tillage, continuous plow tillage also significantly increased dry matter weight of fragrant rice at maturity as well as total N, P, and K accumulation in plant tissues. Moreover, the net income of fragrant rice under plow tillage was US$114.05–526.85 higher than rotary tillage per hectare in the last three cropping seasons. In conclusion, continuous plow tillage enhances fragrant rice performance (yield, quality, aroma) relative to rotary tillage. Our results also improve the scientific database on effects of tillage on rice production. [ABSTRACT FROM AUTHOR]

The elevation detection accuracy of the laser receiver in the laser levelling control system directly affects land-levelling operations. To effectively improve the effect of levelling operations and meet the requirements for the accuracy of elevation detection in different industries, this study optimization designed a multilevel adjustable laser receiver. First, we examined the laser signal detection technology and processing circuit, designed the photoelectric conversion array for the detection of the rotating laser, and converted it into a photocurrent signal. We also designed the filter, amplifier, and shaping and stretching circuits for analogue-to-digital conversion of the photocurrent signal. The digital signal was calculated based on the deviation of the elevation by using a microprocessor and was output by a controller area network (CAN) bus. The laser beam spot diameter transmission and diffusion were then studied, and with the detectable spot diameters were compared and analyzed. Accordingly, an algorithm was proposed to calculate the deviation of laser receiver elevation. The resolution of the elevation deviation was set to ±3 mm; however, this value could be adjusted to ±6 mm, ±9 mm, ±12 mm, and ±15 mm, according to the requirements. Finally, the laser receiver was tested and analyzed, and the test results of the elevation detection accuracy showed that when the laser receiver was within a radius of 90 m, the elevation detection accuracy was within the ±3 mm range. The outcomes of the farmland-levelling test showed that the standard deviation S d of the field surface decreased from 9.54 cm before levelling to 2.42 cm after levelling, and the percentage of sampling points associated with absolute errors of ≤3 cm was 84.06%. These outcomes meet the requirements of high-standard farmland construction. The test results of concrete levelling showed that within a radius of 30 m, the standard deviation S d of the elevation adjustment of the left laser receiver was 1.389 mm, and the standard deviation S d of the elevation adjustment of the right laser receiver was 1.316 mm. Furthermore, the percentage of the sampling points associated with absolute elevation adjustment errors of ≤3 mm in the cases of the two laser receivers was 100% after levelling, whereas the standard deviation S d of the sand bed surface was 0.881 mm. Additionally, the percentage of the sampling points associated with absolute errors of ≤3 mm was 100%. This met the construction standards of the concrete industry. [ABSTRACT FROM AUTHOR]