Search

为实现对市面上常见以及实际案件中出现的动物油样本进行快速无损识别，借助光谱分析技术和机器学习算法，基于连续分类策略，对不同动物油样本在种类及品牌/来源地方面进行区分和认定。收集了247份动物油样本(鸡油、牛油、鹅油、猪油、羊油、鸭油)，对其进行红外光谱扫描，采用自动基线校正和峰面积归一化消除样本基线漂移和量纲不一致情况；再分别采用Savitzky-Golay平滑、二项式平滑、邻域平均法、FFT滤波、一阶导数和二阶导数对红外光谱进行预处理，比较了6种预处理方法在降噪方面的差异性，同时构建不同预处理方法下的随机森林、贝叶斯网络以及最小二乘支持向量机3种分类模型，开展各样本“种类—品牌/来源地”的连续分类工作。结果表明，相较于未预处理模型，经过预处理后，模型的识别能力均有提升，其中采用FFT滤波预处理结合随机森林模型可较好区分6种动物油，其对6种动物油样本品牌/来源地的识别准确率由高到低依次为鸡油、牛油、鹅油、猪油、羊油、鸭油；对实际案件中2份检材进行验证性分析，结果与实际情况相符合。红外光谱结合机器学习算法可应用于基于连续分类策略的动物油的快速无损识别。To achieve rapid and non-destructive recognition of animal oil samples commonly found in the market and in actual cases, different animal oil samples were distinguished and identified in terms of category and brand/origin by spectral analysis technology and machine learning algorithms based on continuous classification strategy.A total of 247 animal oil （chicken oil, beef tallow, goose oil, lard, duck fat, mutton fat） were collected and scanned by infrared spectra, and automatic baseline correction and peak area normalization were used to eliminate the baseline drift and dimensional inconsistency. The infrared spectra of animal oil was pretreated by Savitzky-Golay smoothing, binomial smoothing, neighborhood average, FFT filter, first-order derivative and second-order derivative,respectively,and the differences of the six pretreatment methods in noise reduction were compared. Besides, the random forest, Bayesian network and least square support vector machine models were constructed with six pretreatment methods respectively to carry out the continuous classification of "category—brand/origin" of all samples. The results showed that the recognition ability of models was improved after pretreatment. The random forest model after FFT filter pretreatment could better distinguish six types of animal oil, the classification accuracy of recognizing the brands/origin of the six animal oil samples from high to low was chicken oil, beef tallow, goose oil, lard, mutton fat and duck fat. Confirmatory analysis was made on two samples from the actual case, and the results were consistent with the actual situation. Infrared spectroscopy combined with machine learning algorithms can apply to the quick and non-destructive recognition of animal oil based on continuous classification strategy.

Based on the problem that considerable ballistic error and inaccurate target coverage of terminal-sensitive projectiles while fighting without weather conditions causing the reduction of combat effectiveness and/or even can’t complete combat tasks, this work elaborated two commonly used solutions and methods. According to the real working experience of the test range, a new proposal of using the meteorological forecast data to calculate the ballistic elements of the terminal-sensitive projectile was studied. Furthermore, the deviation between the ballistic landing point and the test landing point under various environmental factors such as standard meteorological conditions, meteorological forecast data and actual meteorological data was analyzed using numerous testing data and ballistic simulation results, which indicates the accuracy and reliability of meteorological forecast data in the terminal sensitive projectiles range and side deflection prediction. In summary, this study proves a theoretical reference for improving the combat effectiveness of terminal-sensitive projectiles under real combat environments, and also provides convenience for the development of terminal-sensitive projectile tests.

Oversized coal blocks can easily cause poor coal flow, blockage, and coal stacking when transported on a belt conveyor. However, the differences in appearance and color between large coal blocks and ordinary coal blocks are small, and there are obstructions and stacking between coal blocks. Existing coal block detection methods are not precise enough to distinguish between large coal blocks and ordinary coal blocks, which can easily lead to missed or false detections. In order to solve the above problems, a modified YOLOv5 model is proposed for detecting large coal blocks in belt conveyors. The model uses parallel dilated convolution modules to replace some ordinary convolution modules in the YOLOv5 backbone network. It expands the receptive field, improves multi-scale feature learning capability, and better distinguishes large coal blocks from ordinary coal blocks. The joint attention module is added to the neck network to better integrate contextual information and improve the positioning capability for large coal blocks. The model uses the trained improved YOLOv5 model to detect real-time coal transportation videos captured by the camera, and links PLC alarms in real-time based on the quantity information of large coal blocks. The experimental results show that compared to the original YOLOv5 model, the improved YOLOv5 model has improved recall and average precision by 3.4% and 2.0%, respectively. PLC can operate corresponding indicator lights and buzzers to alert based on the quantity of large coal blocks detected by the improved YOLOv5 model. The improved YOLOv5 model is applied to actual coal transportation videos in coal mines, with a detection precision of 97.0% for large coal blocks, effectively avoiding missed and false detections.