Automated UAV and satellite image analysis for wildlife monitoring

为了生态/保护的目的，特别是为了地球仪无法到达的地区，如南极洲，人们越来越有兴趣应用无人机和卫星获取的图像来监测野生动物。关于最后位置，英国南极调查局定期收集图像数据。由人类手动分析这些图像是一项繁琐且昂贵的任务，这强烈推动了自动图像处理解决方案的开发。也就是说，据我们所知，现有的算法不提供所需的性能/鲁棒性。该项目旨在开发用于检测和计数野生动物的自动计算机视觉算法。最初，我们将专注于海豹和企鹅的图像，但目的是开发足够通用的方法，可以适合监测其他野生动物，并有可能将这项技术用于生态/保护之外的其他应用。方法学最近，一系列被称为“深度学习”的计算机视觉算法已被报道在许多图像处理/计算机视觉任务中提供性能的逐步变化。在计算机视觉中，深度学习通常使用深度卷积神经网络（CNN）。基于DL和CNN的算法的关键特征是它们用自动分层特征学习代替了现有技术算法中设计手工特征的步骤。作为博士学位的一部分，成功的候选人将研究深度学习算法在相关领域的开发和应用，即计算图像中的野生动物。学生将利用从卫星、有人驾驶飞机和无人机收集的图像捕获的数据。该项目的一个关键方面将是提供关于图像要求的建议，以确保算法稳健性达到所需水平。新开发的算法将与现有技术进行比较。设想的系统将需要大量的注释图像数据集进行训练，这将需要有关对象的图像外观的一些专家知识。学生将使用现有的数据库，但也需要与BAS的相关专家密切联系，以便在必要时扩展这些数据集。培训NEXUSS CDT提供最先进的，高度体验式的培训，用于环境科学的尖端智能和自主观测系统的应用和开发，以及全面的个人和专业发展。学生将有广泛的机会，通过与学术，研究和工业/政府/政策合作伙伴的广泛网络的互动，扩大他们的多学科视野。该学生将在东安格利亚大学注册，在图形，视觉和语音实验室的计算科学学院托管。学生将接受与项目相关的所有领域的培训，包括计算机视觉，机器学习以及Matlab和Python编程。学生将在英国南极调查局工作一段时间，以熟悉该项目的图像和生态方面。F.冈萨雷斯湾a. Montes，E. Puig，S.约翰逊，K. Mengersen和K. J.加斯顿，无人机（无人机）和人工智能革命野生动物监测和保护，传感器2016，16，97; doi：10.3390/s16010097V。Lempitsky和A.塞瑟曼"学习计算图像中的物体。"神经信息处理系统的进展。2010.G.法语，M。H. Fisher，M. Mackiewicz和C.L. Needle，卷积神经网络用于在渔业监控视频中计数鱼类，2015年，第26届英国机器视觉会议上的动物及其行为研讨会

Project RationaleThere is an increasing interest in application of UAVS (Unmanned Aerial Vehicles) and satellite acquired imagery for monitoring wildlife for ecology/conservation purposes including in particular inaccessible areas of the globe such as Antarctic.With regard to the last location, image data are regularly collected by the BritishAntarctic Survey (BAS). The manual analysis of this imagery by humans is a tedious and expensive task which strongly motivates the development of an automated image processing solutions. This said, to our knowledge the existing algorithms do not provide the required performance/robustness. This project will aim to develop automated computer vision algorithms for detection and counting of wildlife. Initially, we will focus on the seal and penguin imagery, but the aim is to develop methods generic enough that could suit monitoring other wildlife with a possibility of using this technology for other applications beyond ecology/conservation.MethodologyRecently, a family of computer vision algorithms known as 'Deep Learning' has been reported to provide a step-change in performance in many image processing/computer vision tasks. In computer vision Deep Learning usually utilizes a deep convolutional neural network (CNN). The key feature of DL and CNN based algorithms is that they replace the step of designing handcrafted features in the prior art algorithms with the automated hierarchical feature learning. As part of their PhD, a successful candidate will investigate development and application of Deep Learning algorithms for the relevant field i.e. counting wildlife in images. The student will make use of data captured using imagery collected from satellites, manned aircraft and UAVs. A key aspect of the project will be to provide the recommendations on the requirements of the imagery allowing for ensuring the required level of algorithm robustness. The new developed algorithms will be compared to the prior-art. The envisaged system will require a large dataset of annotated imagery for training and this will require some expert knowledge on the image appearance of the relevant objects. The student will use the existing databases when available, but will also need to closely liaise with the relevant experts in the BAS for extending those datasets if necessary.TrainingThe NEXUSS CDT provides state-of-the-art, highly experiential training in theapplication and development of cutting-edge Smart and Autonomous ObservingSystems for the environmental sciences, alongside comprehensive personal and professional development. There will be extensive opportunities for students to expand their multi-disciplinary outlook through interactions with a wide network of academic, research and industrial / government / policy partners. The student will be registered at University of East Anglia, hosted at School of Computing Sciences in the Graphics, Vision and Speech laboratory. The student will receive training in all areas relevant to the project including computer vision, machine learning as well as Matlab and Python programming. The student will spend periods of time at British Antarctic Survey in order to familiarize with the images and the ecological aspects of the project.References L. F. Gonzalez, G. A. Montes, E. Puig, S. Johnson, K. Mengersen and K. J.Gaston, Unmanned Aerial Vehicles (UAVs) and Artificial IntelligenceRevolutionizing Wildlife Monitoring and Conservation, Sensors 2016, 16, 97;doi:10.3390/s16010097V. Lempitsky and A. Zisserman. "Learning to count objects in images." Advancesin Neural Information Processing Systems. 2010.G. French, M. H. Fisher, M. Mackiewicz and C.L. Needle, Convolutional NeuralNetworks for Counting Fish in Fisheries Surveillance Video, 2015, Machine Visionof Animals and their Behaviour Workshop at the 26th British Machine VisionConference