Digital image is an important component of multimedia information, it has huge information, and is easy to analyze, process, store and transmit. These are main characters of digital image .Image processing techniques such as edge contour,image enhancement and restoring ,many mathematical transforms and others, have been widely used in many applications , such as medical treatment , satellite remote sensing and so on. The character of the technique is oriented to human visual system and enhance or stress the feature area to reach better effect by the way of analyzing image and extracting features to make those explain the meanings of the image. Because the content of image is complicated, the traditional methods has poor adaptive capacity and low automation, that is to say those methods lack for "intelligence". Under such circumstances, we apply to "the tight coupling and multi-agents image analysis system" and research and develop "intelligence" image analysis system in order to detect cashmere and criminal trace., which is based on the software agent, neural network, misty mathematics and other artificial intelligence techniques.With the finance aid of National Nature Scientific Foundation ,research group has made great effort to gain the following achievements in this area:(1)The method of domain-oriented image segmentation.(2)The method of extracting image geometrical and texture features.(3)The managing method of image analysis data around database.(4)The data mining method of neural network and fuzzy mathematics..(5)In the system structure ,we absorb the idea of software agent technique.Having achieved the above task, the research group has developed an image analysis system based on windows environment using Visual C++ and SQL Server. The input devices of this system are scanner, CCD, microscope, digital camera and others. It has friendly and easy human-computer interface, high adaptive capacity and intelligence. This system has been used in many polices and has received good remarks, the detection of cashmere is now an experiment
作为人工智能方法论的研究,本项目结合山羊绒计算机自动检验这一具体应用,以人工神经网络、软件代理技术为主线条,充分利用面向对象、信息提取、符号推理、基于知识系统、模式识别、数字图像处理等技术,在概念、结构和方法上从事基于图像内容进行分析的研究。其目标在于开发一套一般化机器自动学习、软件间相互协调、基于知识的图像分类系统。
{{i.achievement_title}}
数据更新时间:2023-05-31
多源数据驱动CNN-GRU模型的公交客流量分类预测
涡轮叶片厚壁带肋通道流动与传热性能的预测和优化
不同分子分型乳腺癌的多模态超声特征和临床病理对照研究
简化的滤波器查找表与神经网络联合预失真方法
基于深度神经网络的个体阅读眼动预测
缆系式紧耦合多机器人系统协调建模及稳定性分析
基于模糊层次分析方法的多代理调度/控制系统
多委托人多代理人分销系统的委托代理问题研究
规则驱动的自适应多代理系统研究