基于WMSN的林火监测中图像处理算法的研究
发布时间：2018-08-22 11:10
【摘要】：森林是生态系统中重要的自然资源,森林中火灾的发生会对自然环境、生物多样性和经济安全造成严重威胁。火灾具有蔓延的特性,为避免其对森林造成更大的破坏,开发和利用先进的技术进行林火监测,尽快的发现火情非常重要。无线多媒体传感器网络是可用于林火监测和相关活动的新兴技术,由具备感测、处理和无线通信功能的小型电池供电的传感器节点组成,具有覆盖范围广、成本低的特点,由于森林属于大空间环境且情况复杂,引入图像传感器节点,有助于实现直观、准确的火灾监测。然而传感器节点本身能量有限、计算资源受限,约束了对图像的处理能力,因此,在图像处理技术中如何减少内存的使用、能量的消耗成为WMSN的研究热点与难点。根据无线多媒体传感器网络进行林火监测的实际需要,本文提出了在保证准确率和图像质量前提下的算法,用于减少能量的消耗,主要完成工作如下:(1)基于WMSN的火焰检测方法针对WMSN的资源有限性,森林火灾发生的随机性和小概率性,图像传感器节点采集图像信息量大等问题。本文提出一种基于运动特性和颜色时空特性的火焰检测算法,在节点端直接对图像传感器节点采集到的图像数据进行处理,然后将关注信息传输至网关,从而达到减少网络传输能量的目的。该算法首先利用运动目标检测算法消除静态类火焰颜色物体的干扰,获得运动目标区域以减少后续步骤的处理量;其次通过对火焰序列样本与非火焰序列样本进行I前景区域与S前景区域的相关性分析,得到火焰的颜色时空特性,确定判断火焰的阈值;最后计算当前运动区域的I前景与S前景的相关性系数,将其与阈值进行比较,大于阈值,则认为当前运动区域中含有火焰,将其通过多跳方式传至网关,发送到监测中心,否则,不发送当前运动区域。实验结果表明,该算法能够准确、快速的检测出火焰,且仅发送被确认为火焰的区域到监控中心进行报警,从而减少冗余数据在网络中的传输与能量的消耗。(2)基于WMSN的图像编码方法针对图像数据中含有的大量冗余信息,直接发送图像至网关消耗过多能量的问题,本文在多级树集合分裂算法的基础上,提出一种基于条带缓冲和状态标记图的改进算法,该算法去除图像中的冗余信息后,仅将少量的比特流传输至网关。首先将图像分成多个条带,对每一条带依次进行提升的9/7小波变换,并将其放入条形缓冲区中;其次,将小波系数的重要性标志放入状态标记图中,通过扫描状态标记图,得到输出码流,同时释放条形缓冲区用于存放小波系数;依次类推,直到图像被完整编码,开始下一个图像的处理。实验结果表明,该方法大大减少了网络中冗余信息的传输,并且具有较小的内存占用量与较快的处理速度,有效的减少了能量的消耗。
[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.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S762.32;TP391.41
[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.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S762.32;TP391.41
【参考文献】
相关期刊论文 前10条
1 林婉怡;顾星;殷淑s，
本文编号：2196922
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2196922.html
