基于声谱图纹理特征的蛋鸡发声分类识别

为有效地辨别蛋鸡不同类型声音,了解蛋鸡的健康状况以及个体需求,提高生产效率的同时改善蛋鸡福利化养殖,提出一种基于声谱图纹理特征的蛋鸡发声分类识别方法。以海兰褐蛋鸡的声音为研究对象,将图像处理和声音处理技术相结合,由一维声音信号转换为二维图像信号,二维声谱图中的纹理特征呈现了蛋鸡声音的更多细节信息。最后,利用2DGabor滤波器提取蛋鸡发声声谱图中的声纹信息,并采用人工神经网络模型进行训练和分类识别。试验结果表明,本文方法平均灵敏度和平均精确度不低于92.0%,风机噪声识别灵敏度达99.3%,鸣叫声识别灵敏度最低,为76.0%。

The computational fluid dynamic modeling of downwash flow field for a six-rotor UAV

The downwash flow field of the multi-rotor unmanned aerial vehicle(UAV), formed by propellers during operation, has a significant influence on the deposition, drift and distribution of droplets as well as the spray width of the UAV for plant protection. To study the general characteristics of the distribution of the downwash airflow and simulate the static wind field of multi-rotor UAVs in hovering state, a 3 D full-size physical model of JF01-10 six-rotor plant protection UAV was constructed using Solid Works. The entire flow field surrounding the UAV and the rotation flow fields around the six rotors were established in UG software. The physical model and flow fields were meshed using unstructured tetrahedral elements in ANSYS software.Finally, the downwash flow field of UAV was simulated.With an increased hovering height, the ground effect was reduced and the minimum current velocity increased initially and then decreased. In addition, the spatial proportion of the turbulence occupied decreased. Furthermore, the appropriate operational hovering height for the JF01-10 is considered to be 3 m. These results can be applied to six-rotor plant protection UAVs employed in pesticide spraying and spray width detection.

Short-term feeding behaviour sound classification method for sheep using LSTM networks

A deep learning approach using long-short term memory(LSTM)networks was implemented in this study to classify the sound of short-term feeding behaviour of sheep,including biting,chewing,bolus regurgitation,and rumination chewing.The original acoustic signal was split into sound episodes using an endpoint detection method,where the thresholds of short-term energy and average zero-crossing rate were utilized.A discrete wavelet transform(DWT),Mel-frequency cepstral,and principal-component analysis(PCA)were integrated to extract the dimensionally reduced DWT based Mel-frequency cepstral coefficients(denoted by PW_MFCC)for each sound episode.Then,LSTM networks were employed to train classifiers for sound episode category classification.The performances of the LSTM classifiers with original Mel-frequency cepstral coefficients(MFCC),DWT based MFCC(denoted by W_MFCC),and PW_MFCC as the input feature coefficients were compared.Comparison results demonstrated that the introduction of DWT improved the classifier performance effectively,and PCA reduced the computational overhead without degrading classifier performance.The overall accuracy and comprehensive F1-score of the PW_MFCC based LSTM classifier were 94.97%and 97.41%,respectively.The classifier established in this study provided a foundation for an automatic identification system for sick sheep with abnormal feeding and rumination behaviour pattern.

Extracting body surface dimensions from top-view images of pigs

Continuous live weight and carcass traits estimation are important for the pig production and breeding industry.It is widely known that top-view images of a pig’s body(excluding its head and neck)reveal surface dimension parameters,which are correlated with live weight and carcass traits.However,because a pig is not constrained when an image is captured,the body does not always have a straight posture.This creates a big challenge when extracting the body surface dimension parameters,and consequently the live weight and carcass traits estimation has a high level of uncertainty.The primary goal of this study is to propose an algorithm to automatically extract pig body surface dimension parameters,with a better accuracy,from top-view pig images.Firstly,the backbone line of a pig was extracted.Secondly,lengths of line segments perpendicular to the backbone line were calculated,and then feature points on the pig’s contour line were extracted based on the lengths variation of the perpendicular line segments.Thirdly,the head and neck of the pig were removed from the pig’s contour by an ellipse.Finally,four length and one area parameters were calculated.The proposed algorithm was implemented in Matlab®(R2012b)and applied to 126 depth images of pigs.Taking the results of the manual labeling tool as the gold standard,the length and area parameters could be obtained by the proposed algorithm with an accuracy of 97.71%(SE=1.64%)and 97.06%(SE=1.82%),respectively.These parameters can be used to improve pig live weight and carcass traits estimation accuracy in the future work.