论文部分内容阅读
本文提出了一种基于图像识别的水质浊度测量方法;即通过CCD捕获水下光源的光散射信息来评价水质浊度信息;提出了基于空间域(即梯度)和频率域(即小波变换)的图像信息提取算法,采用反向传输方式的神经网络对水质浊度的图像信息进行分类和标定。神经网络测试结果表明,本系统的平均误差率为0.5%,最大误差为3%,从性能上可以满足大量程浊度测量的需求。
In this paper, a method of water turbidity measurement based on image recognition is proposed. The light scattering information of underwater light source is captured by CCD to evaluate the turbidity information of water quality. Based on the spatial domain (ie gradient) and frequency domain (ie wavelet transform) Image information extraction algorithm, the use of reverse transmission neural network to classify and calibrate the image information of water turbidity. Neural network test results show that the average error rate of the system is 0.5%, the maximum error of 3%, from the performance to meet the needs of large turbidity measurement.