Transactions of the Chinese Society of Agricultural Engineering, the 1st in Agricultural Engineering, is supervised by China Association for Science and Technology, and sponsored by Chinese Society of Agricultural Engineering. It aims to introduce the latest scientific achievements and developing trends of Agricultural Engineering and provides the academic developments abroad and domestic of the discipline. The scope covers agricultural water-soil engineering, agricultural information and electrical technology, agricultural products processing engineering.
The journal is included in EI, JST, Pж(AJ), CA and CSCD.
Editor-in-Chief Zhu Ming
Deputy Editor-in-Chief Wei Xiuju Zhang Ruihong Xi Weimin Wang Liu Wang Yingkuan Li Pingping Ying Yibin Tong Jin Yun Wenju Zhao Chunjiang Kang Shaozhong
Before fruit thinning, the factors such as complex background, various illumination conditions, foliage occlusion, fruit clustering, especially the extreme similarities between apples and background, made the recognition of small apple targets very difficult. To solve these problems, we proposed a region-based fully convolutional network (R-FCN) recognition method. Firstly, the deep convolution neural network including ResNet-50 based on R-FCN and ResNet-101 based on R-FCN were studied and analyzed. After the framework of these two networks was analyzed, it was obvious that the difference between these two networks was the ‘conv4’ block. The ‘conv4’ block of ResNet-101 based on R-FCN was 51 more layers than that of ResNet-50 based on R-FCN, but the recognition accuracy of these two networks was almost the same. By comparing the framework and recognition result of ResNet-50 based on R-FCN and ResNet-101 based on R-FCN, we designed the R-FCN based on ResNet-44 to improve the recognition accuracy and simplify the network. The main operation to simplify the network was to simplify the ‘conv4’ block, and the ‘conv4’ block of ResNet-44 based on R-FCN was six layers fewer than that of ResNet-50 based on R-FCN. The Res Net-44 based on R-FCN consisted of ResNet-44 fully convolutional network (FCN), region proposal network (RPN) and region of interest (RoI) sub-network. Res Net-44 FCN, the backbone network of R-FCN, was used to extract the features of the image. The features were then used by RPN to generate RoIs. After that, the features extracted by ResNet-44 FCN and RoIs generated by RPN were used by RoI sub-network to recognize and locate small apple targets. A total of 3 165 images were captured in an experimental apple orchard in the College of Horticulture, Northwest A&F University, in Yangling City, China. After image resizing and manual annotation, 332 images, including 85 images captured under sunny direct sunlight condition, 88 images captured under sunny backlight condition, 86 images captured under cloudy direct sunlight condition, 74 images captured under cloudy backlight condition, were selected as test set, and the other 2 833 images were used to train and optimize the network. To enrich image training set, the data augment, including brightness enhancement and reduction, chroma enhancement and reduction, contrast enhancement and reduction, sharpness enhancement and reduction, and adding Gaussian noise, was performed, then a total of 28 330 images were obtained with 23 591 images randomly selected as the training sets, and the other 4 739 images as validation sets. After training, the simplified ResNet-44 based on R-FCN was tested on the test set, and the experimental results indicate that the method can effectively apply to the images captured under different illumination conditions. The method can recognize clustering apples, occluded apples, vague apples and apples with shadows, strong illumination and weak illumination on the surface. In addition, the apples divided into parts by branched or petiole cloud can also be recognized effectively. Overall, the recognition recall rate can achieve 85.7%. The recognition accuracy and false recognition rate are 95.2% and 4.9%, respectively. The average recognition time is 0.187 s per image. To further test the performance of the proposed method, we compared the other three methods, including Faster R-CNN, ResNet-50 based on R-FCN and ResNet-101 based on R-FCN. The
F1 of the proposed method is increased by 16.4, 0.7 and 0.7 percentage points, respectively. The average running time of the proposed method improves by 0.010 s and 0.041 s compared with that of Res Net-50 based on R-FCN and Res Net-101 based on R-FCN, respectively. The proposed method can achieve the recognition of small apple targets before fruits thinning which cannot be realized by traditional methods. It can also be widely applied to the recognition of other small targets whose features are similar to the background.
Flooding is one of main disasters for agricultural production in China. Its influence on rice is rarely studied in Guizhong District in Guangxi. In this study, we investigated the effects of different flooding durations at different growth stages on the growth and yield components of rice. The experiment was carried out in Nalu Village of Laibing City, Guangxi, China (109°52’ E, 24°4’N). The soil was sandy loam. As early and late rice, Sixiang 1 and Baixiang 139 were planted in pots in 2015 and 2016, respectively. Flooding of early rice started at different growth stages of tillering, booting and flowering. At each growth stage, flooding lasted for 0 (control), 2, 4, 6, 8, and 10 days, respectively. Because the precipitation during the booting and flowering stages of late rice was small, the flooding of late rice was carried out on three different days at tillering stage. At flooding, the water surface remained 2–3 cm higher than the plant height. The flooding duration of late rice was same as that of early rice. During the experiment, the tiller numbers before and after flooding, the heading stage time and its duration, grain number per panicle, seed-setting rate, 1 000-grain weight and yield per pot after harvest were determined, and the relationships between yield and yield components were subjected to path analysis. The results showed that the seedling death rate was less than 7.5% when the flooding duration at the tillering stage was less than 2 days, which was not significantly different from control. As the flooding duration increased to 4–10 days, the death rate increased. When the flooding lasted more than 6 days, the death rate of early rice could reach more than 80% and even 100%. The death rate of Sixiang 1 reached 50% after flooding for 6 days while it was still smaller than 50% for Baixiang 139 after flooding for less than 10 days. The death rate was below 40% for the treatment of flooding at flowering stage. The initial heading stage time of flooding for 8 days was postponed compared with that of the control when flooding at different growth stages, and the heading duration was also prolonged. The yield was not significantly decreased when the flooding at the tillering stage lasted less than 4 days, but flooding for 6 days could result in the reduction of yield by 80% and flooding for 8–10 days could result in the death of plants. The yield reduction rate of flooding for 2 days at the booting and flowering stages was higher than 50%. The path analysis of yield components showed that the yield was mainly affected by effective panicles, seed-setting rate and 1 000-grain weight for the flooding at the tillering stage, by seed-setting rate, number of grains per panicle and 1 000-grain weight for the flooding at the booting stage, and by effective panicles and seed-setting rate for the flooding at the flowering stage, respectively. The study provides important information for formulating the flooding disaster warning system and management methods of double-cropping rice in Southwest China.
In the development of precision agriculture, intelligent agricultural machinery is an effective way to alleviate the current tense situation in world food security. The complex field environment and meticulous work effects require the agricultural machinery to have the ability of perceiving the attitude of the agricultural machinery accurately in real time. For example, the precision navigation control and the leveling control of agricultural machinery are all dependent on the accurate attitude measurement. What is more, the attitude of agricultural machinery is one of the key parameters of agricultural mechanics modeling and agricultural machinery safety warning learning. However, the external acceleration of the vehicle, which is commonplace under dynamic operation conditions, poses a challenge. The paper developed a minimal hardware system for external acceleration identification and attitude estimation used for field working vehicles and verified by the experiments taken place on the Innova 2100 shaker and the ZP9500 sprayer of high ground clearance in the field in order to further improve the precision operation of agricultural machinery. The modern micro-electromechanical systems (MEMS) technology provides the moderate-cost and miniaturized solutions for the development of attitude reference system. With the highly integrated inertial measurement units (IMUs) ADIS16445 provided by ADI company and micro ARM processor STM32 F446 provided by ST Company, the hardware platform was built. ADIS16445 ISensor® included tri-axial gyroscopes and tri-axial accelerometers. The raw sensors data were sampled by STM32 F446 RC processor through SPI interface. The attitude calculation was carried out based on the direction cosine matrix algorithm. Based on the gyroscope and accelerometer measurement model, a first-order external acceleration measurement model was proposed, and a Kalman filtering fusion algorithm with four state vectors was established. Since the bias of gyroscopes and accelerometers was stable after hardware pre-heating, the impact of bias of MEMS inertial sensors in the fusion algorithm was not considered. Innova 2100 shaker is standard equipment of rotary motion, and different centripetal accelerations can be achieved to verify the external acceleration identification at varying rotation speeds. Innova 2100 shaker test results show that the measurement error is less than 0.214 m/s
2 under the external acceleration lower than 10
g. Field experiments were conducted on Innova 2100 shaker and the ZP9500 sprayer of high ground clearance provided by LOVOL Company with the assistance of attitude and heading reference system (AHRS) MTi300 provided by Xsens Company. The MTi300 AHRS provides high precision attitude and heading output with high stability and fast dynamic response with dynamic measurement accuracy of 0.3°, which makes it widely used in navigation implements, automobiles, agricultural machinery, and other fields. The experiment results from the sprayer of high ground clearance show that compared to the MTi300 AHRS, the average and maximum measurement errors of the roll angle are 0.069° and 0.23°, respectively. The average and maximum measurement errors of pitch angle are 0.078° and 0.39°, respectively. The test results verify that the proposed Kalman filtering algorithm is accurate and stable, which can improve the quality of agricultural machinery operations and have more applicability.