Use of chemical cues in detection of conspecific predators and prey by newts, <Emphasis Type="Italic">Notophthalmus viridescens</Emphasis>

Summary. Detection of heterospecific predators and prey via chemical cues is well known, but only a few studies have examined the potential for such discrimination in cannibalistic systems. In newts, Notophthalmus viridescens, adults are opportunistic predators of conspecific larvae. I used a laboratory bioassay to determine whether larvae and adults distinguish between chemical stimuli from members of the different age classes. Larvae distinguished between chemical stimuli from larvae and adults by decreasing their activity only following exposure to stimuli from adults. Decreased activity is consistent with an antipredator response in many prey species, including larval newts. In contrast, adults increased their activity and increased time spent in open areas in response to stimuli from larvae, but not to stimuli from adults. Increased activity is consistent with a feeding response; adults also showed increased activity and increased time in open areas in response to chemical stimuli from familiar heterospecific prey (brine shrimp). The proximate cue that allowed the newts to distinguish between the different age classes is not associated with short-term dietary differences because all stimulus animals were fed the same prey. Stimulus strength was controlled by diluting the stimulus solutions according to the volume of the stimulus animal. Therefore, there appear to be intrinsic differences in the chemical signatures of larval and adult newts.     

基于监督局部映射的煤矿井下人员身份鉴别方法

为提高煤矿井下人员身份识别率,在局部保持投影(LPP)算法的基础上,提出监督局部映射(SLP)算法。该方法充分利用数据的局部和非局部信息及类别信息,对数据进行维数约简,使特征空间同类数据间的距离更小,不同类数据间的距离更大。该方法能够克服煤矿井下艰苦、空间受限环境中人脸、虹膜和指纹识别率不高的问题。在真实步态数据库上的实验结果表明,基于步态的煤矿井下人员身份鉴别是可行的。     

模式识别在火灾调查中的汽油分类问题的应用研究

在火灾调查中,检测汽油成分并对其进行正确分类尤为重要.运用GC-MS对90#和93#两种普通汽油的共50个样本进行检测,所得的GC-MS原始数据通过PCA方法进行处理,以提取有用信息,避免冗余变量进入后续计算.在此基础上应用KNN方法对这两种汽油助燃剂进行分类.结果表明, KNN方法对这两种汽油的分类准确率达到100%,且当初始数据未经标准化预处理时也能达到同样准确的分类效果.研究表明:将模式识别方法正确地运用到助燃剂鉴定和分类工作中有助于火灾调查.     

Discrimination of different social companions in spectacled parrotlets (Forpus conspicillatus): evidence for individual vocal recognition

Individual recognition is generally assumed to be a prerequisite for establishing and maintaining a complex social system. Indeed, there is good evidence that highly social species have complex systems of vocal communication with individual recognition by acoustic cues. In this study, we provide experimental evidence that vocal class and individual recognition is present in a non-passerine bird, the spectacled parrotlet (Forpus conspicillatus). Spectacled parrotlets live in a complex system of social relationships. Soon after fledging, the young establish close sibling relationships which are important for successful socialization, pairing and reproduction. In a series of playback experiments we tested if spectacled parrotlets use contact calls for vocal recognition. The results showed that spectacled parrotlets discriminate between the contact calls of different social categories. Adult birds preferred to respond to the contact calls of their mates. Subadult individuals recognized the contact calls of their siblings. During the period of pair bond formation, the affiliative contacts to the siblings decrease, but the parrotlets continue to respond to the calls of their siblings. This is the first evidence that vocal sibling recognition might outlast the period of strong sibling interaction and extends into the period of pair bond formation. In cases of mate loss or divorce, the acoustic contact to their siblings might facilitate the re-establishment of close sibling relationships. Received: 29 October 1997 / Accepted after revision: 5 April 1998     

虹膜识别在肉类食品安全追溯系统中的应用及关键技术研究

为了提高大型动物肉类食品的安全性,在研究肉类食品供应链的安全管理的基础上,将大型动物的虹膜识别技术引入食品供应链管理中,构建基于虹膜识别的大型动物肉类食品安全可追溯系统,达到肉类食品安全信息查询及出现安全问题后的个体溯源的目的。借鉴人眼虹膜识别的相关技术,以牛眼为例对大型动物虹膜识别的相关算法和虹膜定位等关键技术进行研究,运用改进的Sobel算子对图像进行边缘检测;引入二次B样条曲线算法实现了牛眼虹膜的精确定位;分析牛虹膜纹理特征的分布特点,实现了虹膜特征区域的提取及归一化处理。研究成果对大型动物的虹膜识别研究和肉类食品供应链安全管理具有一定的参考意义。     

基于FTA-SVM的车辆发动机故障识别方法

发动机结构日益复杂,其故障具有多样性和频发性的特点,收集大量故障样本存在很多实施障碍。为了提高车辆发动机的故障识别的效率和准确性,提出了一种新的结合故障树(FTA)和支持向量机(SVM)各自特点,从故障模式分析到故障类型识别的FTA-SVM故障识别方法。首先利用故障树在复杂系统故障模式分析中的优势,找出系统的故障模式,建立故障树模型,通过对故障树模型中各故障事件的分析,采集与故障事件状态相关的数据,建立数据与故障树底事件的映射模型,最后利用支持向量机在小样本数据处理中的优势,进行故障类型的识别。以发动机的失火故障为例建立了发动机失火故障树模型及故障数据与故障模式映射模型,验证了FTA-SVM方法的有效性和适用性。     

A dual-sensor-based method to recognize pipeline leakage and interference signals

During the detection of pipeline leakages, false alarms of leak detection could be markedly reduced if the interference signals resulting from pressure regulating, pump regulating or valve movements could be accurately distinguished. A digital recognition method for interference signals and leakage signals based on a dual-sensor system is proposed in this paper. It is demonstrated that the direction of the signal can be recognized by a cross-correlation calculation between two signals from the dual-sensor, one of which undergoes forward linear interpolation and backward linear interpolation. Based on this theory, the interference signal and the leak signal can be discriminated exactly, and the distance between the two sensors in the dual-sensor system can be considerably reduced without needing to increase the sampling frequency. The monotonicity of the cross-correlation function is demonstrated, and a fast discrimination algorithm based on a binary extreme search method, which decreases the computational load and maintains global optimization, is also proposed. A pre-processing method of the actual signal is proposed to decrease the identity requirement for the two sensors in a dual-sensor system. In the experiment based on artificial signals, the proposed discrimination algorithm could achieve accurate recognition of the abnormal signal, and as such, the theory and application of pipeline leak detection based on dual-sensor systems are extended.     

基于BERT-BILSTM-CRF模型的电力行业事故文本智能分析*

为解决电力行业事故报告文本较长、语义复杂,难以进行有效文本识别问题,提出1种以BERT作为底层的预训练模型,并设计1种双重注意力机制编码器,结合BILSTM-CRF深度挖掘事故文本语义特征,从而实现文本智能分析。首先构建电力词典,通过对BERT预训练,进行BIO标注,然后引入BILSTM-CRF模型实现对文本标签智能分类,最后将该模型与现行其他4种深度学习模型进行对比。研究结果表明:该模型智能识别精确率、召回率及F1值(查准率)均达到约97%,较其他4种模型中效果最好的模型分别提高0.02,0.03,0.02。研究结果可为电力行业事故报告文本分析提供1种新思路。     

A novel approach for VOC source apportionment combining characteristic factor and pattern recognition technology in a Chinese industrial area

Volatile organic compound (VOC) emission control and source apportionment in small-scale industrial areas have become key topics of air pollution control in China. This study proposed a novel characteristic factor and pattern recognition (CF-PR) model for VOC source apportionment based on the similarity of characteristic factors between sources and receptors. A simulation was carried out in a typical industrial area with the CF-PR model involving simulated receptor samples. Refined and accurate source profiles were constructed through in situ sampling and analysis, covering rubber, chemicals, coating, electronics, plastics, printing, incubation and medical treatment industries. Characteristic factors of n-undecane, styrene, o-xylene and propane were identified. The source apportionment simulation results indicated that the predicted contribution rate was basically consistent with the real contribution rate. Compared to traditional receptor models, this method achieves notable advantages in terms of refinement and timeliness at similar accuracy, which is more suitable for VOC source identification and apportionment in small-scale industrial areas.     

基于视频识别的直接作业环节安全管控

概述了石化行业现场直接作业环节的安全管控现状,分析了现场管控过程中存在的问题。考虑现场安全管理和监督监控的实际需要,设计了石化行业直接作业现场智能化管控系统,并深入研究作业现场违章行为图像识别的关键技术。现场应用表明,系统充分体现了日常管理、实时监控、智能报警、统计分析等方面的便利性和有效性,各类违章行为图像识别准确率达到80%以上,具有一定的推广价值。