發(fā)布時(shí)間：2018-08-22 11:10

【摘要】：森林是生態(tài)系統(tǒng)中重要的自然資源,森林中火災(zāi)的發(fā)生會(huì)對(duì)自然環(huán)境、生物多樣性和經(jīng)濟(jì)安全造成嚴(yán)重威脅。火災(zāi)具有蔓延的特性,為避免其對(duì)森林造成更大的破壞,開發(fā)和利用先進(jìn)的技術(shù)進(jìn)行林火監(jiān)測(cè),盡快的發(fā)現(xiàn)火情非常重要。無(wú)線多媒體傳感器網(wǎng)絡(luò)是可用于林火監(jiān)測(cè)和相關(guān)活動(dòng)的新興技術(shù),由具備感測(cè)、處理和無(wú)線通信功能的小型電池供電的傳感器節(jié)點(diǎn)組成,具有覆蓋范圍廣、成本低的特點(diǎn),由于森林屬于大空間環(huán)境且情況復(fù)雜,引入圖像傳感器節(jié)點(diǎn),有助于實(shí)現(xiàn)直觀、準(zhǔn)確的火災(zāi)監(jiān)測(cè)。然而傳感器節(jié)點(diǎn)本身能量有限、計(jì)算資源受限,約束了對(duì)圖像的處理能力,因此,在圖像處理技術(shù)中如何減少內(nèi)存的使用、能量的消耗成為WMSN的研究熱點(diǎn)與難點(diǎn)。根據(jù)無(wú)線多媒體傳感器網(wǎng)絡(luò)進(jìn)行林火監(jiān)測(cè)的實(shí)際需要,本文提出了在保證準(zhǔn)確率和圖像質(zhì)量前提下的算法,用于減少能量的消耗,主要完成工作如下:(1)基于WMSN的火焰檢測(cè)方法針對(duì)WMSN的資源有限性,森林火災(zāi)發(fā)生的隨機(jī)性和小概率性,圖像傳感器節(jié)點(diǎn)采集圖像信息量大等問題。本文提出一種基于運(yùn)動(dòng)特性和顏色時(shí)空特性的火焰檢測(cè)算法,在節(jié)點(diǎn)端直接對(duì)圖像傳感器節(jié)點(diǎn)采集到的圖像數(shù)據(jù)進(jìn)行處理,然后將關(guān)注信息傳輸至網(wǎng)關(guān),從而達(dá)到減少網(wǎng)絡(luò)傳輸能量的目的。該算法首先利用運(yùn)動(dòng)目標(biāo)檢測(cè)算法消除靜態(tài)類火焰顏色物體的干擾,獲得運(yùn)動(dòng)目標(biāo)區(qū)域以減少后續(xù)步驟的處理量;其次通過對(duì)火焰序列樣本與非火焰序列樣本進(jìn)行I前景區(qū)域與S前景區(qū)域的相關(guān)性分析,得到火焰的顏色時(shí)空特性,確定判斷火焰的閾值;最后計(jì)算當(dāng)前運(yùn)動(dòng)區(qū)域的I前景與S前景的相關(guān)性系數(shù),將其與閾值進(jìn)行比較,大于閾值,則認(rèn)為當(dāng)前運(yùn)動(dòng)區(qū)域中含有火焰,將其通過多跳方式傳至網(wǎng)關(guān),發(fā)送到監(jiān)測(cè)中心,否則,不發(fā)送當(dāng)前運(yùn)動(dòng)區(qū)域。實(shí)驗(yàn)結(jié)果表明,該算法能夠準(zhǔn)確、快速的檢測(cè)出火焰,且僅發(fā)送被確認(rèn)為火焰的區(qū)域到監(jiān)控中心進(jìn)行報(bào)警,從而減少冗余數(shù)據(jù)在網(wǎng)絡(luò)中的傳輸與能量的消耗。(2)基于WMSN的圖像編碼方法針對(duì)圖像數(shù)據(jù)中含有的大量冗余信息,直接發(fā)送圖像至網(wǎng)關(guān)消耗過多能量的問題,本文在多級(jí)樹集合分裂算法的基礎(chǔ)上,提出一種基于條帶緩沖和狀態(tài)標(biāo)記圖的改進(jìn)算法,該算法去除圖像中的冗余信息后,僅將少量的比特流傳輸至網(wǎng)關(guān)。首先將圖像分成多個(gè)條帶,對(duì)每一條帶依次進(jìn)行提升的9/7小波變換,并將其放入條形緩沖區(qū)中;其次,將小波系數(shù)的重要性標(biāo)志放入狀態(tài)標(biāo)記圖中,通過掃描狀態(tài)標(biāo)記圖,得到輸出碼流,同時(shí)釋放條形緩沖區(qū)用于存放小波系數(shù);依次類推,直到圖像被完整編碼,開始下一個(gè)圖像的處理。實(shí)驗(yàn)結(jié)果表明,該方法大大減少了網(wǎng)絡(luò)中冗余信息的傳輸,并且具有較小的內(nèi)存占用量與較快的處理速度,有效的減少了能量的消耗。
[Abstract]:Forest is an important natural resource in the ecosystem. The fire in the forest will pose a serious threat to the natural environment, biodiversity and economic security. Fire has the characteristics of spreading. In order to avoid causing more damage to the forest, it is very important to develop and use advanced technology to monitor the forest fire and find the fire as soon as possible. Wireless multimedia sensor network is a new technology that can be used in forest fire monitoring and related activities. It is composed of sensor nodes with sensing, processing and wireless communication functions. It has the characteristics of wide coverage and low cost. Because the forest belongs to the large space environment and the situation is complex, the image sensor node is introduced, which is helpful to realize the intuitionistic and accurate fire monitoring. However, the sensor nodes have limited energy and limited computing resources, which restrict the ability of image processing. Therefore, how to reduce the use of memory in image processing technology, energy consumption has become the focus and difficulty of WMSN research. According to the actual needs of wireless multimedia sensor network for forest fire monitoring, this paper proposes an algorithm to reduce energy consumption under the premise of ensuring accuracy and image quality. The main works are as follows: (1) the flame detection method based on WMSN is aimed at the limitation of WMSN resources, the randomness and small probability of forest fire, and the large amount of image information collected by image sensor nodes. In this paper, a flame detection algorithm based on motion characteristics and color space-time characteristics is proposed. The image data collected by the image sensor node is processed directly at the node end, and then the attention information is transmitted to the gateway. In order to achieve the goal of reducing network transmission energy. Firstly, the moving target detection algorithm is used to eliminate the static flame-like color object interference and obtain the moving target area to reduce the processing capacity of the subsequent steps. Secondly, by analyzing the correlation between I foreground region and S foreground region of flame sequence sample and non flame sequence sample, the color space-time characteristic of flame is obtained, and the threshold value of judging flame is determined. Finally, the correlation coefficient between the I foreground and S foreground of the current moving region is calculated and compared with the threshold value. If the threshold is greater than the threshold value, it is considered that there is flame in the current moving region, which is transmitted to the gateway through multi-hop mode and sent to the monitoring center. Otherwise, the current motion area is not sent. The experimental results show that the algorithm can detect the flame accurately and quickly, and only send the confirmed flame area to the monitoring center for alarm. In order to reduce the transmission and energy consumption of redundant data in the network. (2) the image coding method based on WMSN directly sends the image to the gateway and consumes too much energy in view of a large amount of redundant information contained in the image data. Based on the multilevel tree set splitting algorithm, an improved algorithm based on strip buffering and state marking graph is proposed in this paper. After removing redundant information from the image, only a small number of bit streams are transferred to the gateway. Firstly, the image is divided into multiple bands, and each band is transformed by 9 / 7 wavelet transform, which is then put into the bar buffer zone. Secondly, the importance of wavelet coefficients is put into the state marker map, and the state marking map is scanned. The output bit stream is obtained and the bar buffer is released to store the wavelet coefficients, and so on until the image is fully coded to start the next image processing. The experimental results show that the proposed method can greatly reduce the transmission of redundant information in the network, and has smaller memory footprint and faster processing speed, thus effectively reducing the energy consumption.
【學(xué)位授予單位】：太原理工大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：S762.32;TP391.41

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 林婉怡;顧星;殷淑s，

本文編號(hào)：2196922