【摘要】：近年來(lái),隨著城市化進(jìn)程的加快,交通網(wǎng)絡(luò)越來(lái)越復(fù)雜。隨之而來(lái)的追尾碰撞事故發(fā)生頻率越來(lái)越高,然而如何有效地檢測(cè)追尾碰撞事故的發(fā)生成為當(dāng)前智能交通領(lǐng)域急需解決的問(wèn)題。本文試圖用信號(hào)檢測(cè)的方法來(lái)攻克此問(wèn)題。本文提出了一套追尾碰撞檢測(cè)方案來(lái)檢測(cè)追尾事件的發(fā)生。追尾碰撞檢測(cè)方案分為兩部分:追尾碰撞模型和信號(hào)檢測(cè)模型。追尾碰撞模型用來(lái)檢測(cè)當(dāng)前車(chē)輛與后方車(chē)輛是否存在追尾的可能。如果發(fā)現(xiàn)危險(xiǎn)情況,就按照既定的信號(hào)模型一直產(chǎn)生報(bào)警信號(hào)直到危險(xiǎn)情況解除。而后方車(chē)輛的信號(hào)檢測(cè)模型一直在檢測(cè)前方是否有車(chē)輛發(fā)出報(bào)警信號(hào),如果檢測(cè)到報(bào)警信號(hào),就立刻通知司機(jī)進(jìn)行制動(dòng)來(lái)降速或者停車(chē)。從而將目標(biāo)問(wèn)題轉(zhuǎn)化為如何高性能地檢測(cè)報(bào)警信號(hào)。針對(duì)于該問(wèn)題,本文提出兩種信號(hào)檢測(cè)模型:空間相關(guān)信號(hào)檢測(cè)模型和時(shí)間序列檢測(cè)模型。兩種檢測(cè)模型分別利用采集樣本的空間相關(guān)性和時(shí)間相關(guān)性,提高其檢測(cè)概率,增加等檢測(cè)概率條件下的檢測(cè)距離,給后車(chē)司機(jī)更多的反應(yīng)時(shí)間。論文還提出檢測(cè)性能指標(biāo)來(lái)評(píng)價(jià)信號(hào)檢測(cè)模型的性能優(yōu)劣�？臻g相關(guān)信號(hào)檢測(cè)模型利用空間相關(guān)性構(gòu)建信號(hào)檢測(cè)器,來(lái)處理信號(hào)樣本。本文考察了能量檢測(cè)器(Energy detector,ED),能量聯(lián)合檢測(cè)器(AND規(guī)則、OR規(guī)則)以及協(xié)方差檢測(cè)器(Covariance detector,CD),推導(dǎo)了這些檢測(cè)器的檢驗(yàn)統(tǒng)計(jì)量、檢測(cè)門(mén)限值和檢測(cè)概率公式。理論分析比較了能量聯(lián)合檢測(cè)器(AND和OR規(guī)則)和協(xié)方差檢測(cè)器的性能。通過(guò)模擬實(shí)驗(yàn)驗(yàn)證了理論分析的結(jié)論:協(xié)方差檢測(cè)器的檢測(cè)性能比能量聯(lián)合檢測(cè)器(AND、OR規(guī)則)高出50%以上。時(shí)間序列檢測(cè)模型利用時(shí)間相關(guān)性構(gòu)建時(shí)間序列檢測(cè)器,經(jīng)過(guò)時(shí)間序列檢測(cè)器來(lái)處理采樣樣本得到檢驗(yàn)統(tǒng)計(jì)量。本文設(shè)計(jì)了三種時(shí)間序列檢測(cè)器:邏輯聯(lián)合檢測(cè)器、比值聯(lián)合檢測(cè)器以及仲裁聯(lián)合檢測(cè)器。推導(dǎo)了每種檢測(cè)器的檢驗(yàn)統(tǒng)計(jì)量、檢測(cè)門(mén)限值公式。通過(guò)模擬實(shí)驗(yàn)分別對(duì)三種檢測(cè)器進(jìn)行檢測(cè)性能進(jìn)行比較和分析,得出結(jié)論:在判決周期中存在三個(gè)時(shí)間序列時(shí),邏輯聯(lián)合檢測(cè)器的檢測(cè)性能最好;比值聯(lián)合檢測(cè)器次之;仲裁聯(lián)合檢測(cè)器最壞。結(jié)合信號(hào)檢測(cè)模型的檢測(cè)性能指標(biāo),將兩種檢測(cè)模型的信號(hào)檢測(cè)器進(jìn)行性能比較。并得出結(jié)論:1)利用空間相關(guān)性和時(shí)間相關(guān)性可提高信號(hào)檢測(cè)模型的檢測(cè)概率,增加信號(hào)檢測(cè)模型的檢測(cè)距離;2)邏輯聯(lián)合檢測(cè)器的檢測(cè)性能最好。
[Abstract]:In recent years, with the acceleration of urbanization, the traffic network is becoming more and more complex. The frequency of rear-end collision is becoming higher and higher. However, how to effectively detect the rear-end collision has become an urgent problem in the field of intelligent transportation. This paper attempts to solve this problem by signal detection. In this paper, a rear-end collision detection scheme is proposed to detect the occurrence of rear-end events. The rear-end collision detection scheme is divided into two parts: the rear-end collision model and the signal detection model. The rear-end collision model is used to detect whether there is a possibility of rear-end between the current vehicle and the rear vehicle. If a dangerous situation is found, follow the established signal model to generate an alarm signal until the danger is removed. The signal detection model of the rear vehicle always detects whether there is an alarm signal in front of the vehicle. If the alarm signal is detected, the driver is immediately informed to brake to slow down or stop. Thus, the target problem is transformed into how to detect the alarm signal with high performance. To solve this problem, two signal detection models are proposed in this paper: spatial correlation signal detection model and time series detection model. The two detection models improve the detection probability and the detection distance under the condition of equal detection probability by using the spatial correlation and time correlation of the collected samples respectively and give the driver more reaction time. The performance index is also proposed to evaluate the performance of the signal detection model. Spatial correlation signal detection model uses spatial correlation to construct signal detector to process signal samples. In this paper, the Energy detector (Ed), the energy joint detector (AND rule OR rule) and the Covariance detector (Covariance detector CD) are investigated. The test statistics, detection threshold values and detection probability formulas of these detectors are derived. The performance of energy joint detector (AND and OR rule) and covariance detector are analyzed and compared theoretically. The conclusion of the theoretical analysis is verified by simulation experiments: the detection performance of the covariance detector is more than 50% higher than that of the energy joint detector (ANDOR rule). Time series detection model uses time correlation to construct time series detector, which is used to process samples to obtain test statistics. In this paper, three kinds of time series detectors are designed: logic joint detector, ratio joint detector and arbitration joint detector. The test statistics and detection threshold formula of each detector are derived. By comparing and analyzing the detection performance of three kinds of detectors by simulation experiments, it is concluded that when there are three time series in the decision period, the detection performance of the logic joint detector is the best, the ratio joint detector is the second. The arbitration joint detector is the worst. Combining the detection performance index of the signal detection model, the performance of the two detection models is compared. It is concluded that the detection probability of signal detection model can be improved by using spatial correlation and temporal correlation, and the detection distance of signal detection model can be increased. 2) the detection performance of logic joint detector is the best.
【學(xué)位授予單位】：電子科技大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2015
【分類(lèi)號(hào)】：U495
，

本文編號(hào)：2191838