排种器种子流机器视觉图像的滤波及评价

Research on Seeding Image Processing with Different Filtering Methods

在排种器试验及视觉检测过程中,排种器机器视觉图像的滤波是消除信号污染并可靠提取有用信息的必要步骤。采用比较法研究了中值滤波、线性滤波、维纳滤波、小波去噪、中值小波组合滤波等方法在不同滤波器参数下的滤波效果。结果表明,中值滤波效果较好,根据成像状况合理调节滤波器参数,在模板7×7时信噪比可达29.063 1,正确处理率则达0.996 9;中值小波组合滤波法略优于单纯使用中值滤波或小波去噪,其信噪比达到29.165 2,正确处理率则达0.997 1;处理较低光照条件下采集的排种图像,三种方法均达到较理想效果。

The image filtering is anecessary step to avoid noise pollution and distill the useful information in the process of seeding box by machine vision. Attempts to change parameter of median filter, linear filter, Venus filter, wavelet transform and the combination filter of median filter and wavelet transform to processing seeding image were studied. Results indicate that the Signal Noise Ratio is 29. 063 1 and the Correct Process Ratio is 0. 996 9 by median filtering method. The combination filter method is slightly better than simple use with one filter method. The Signal Noise Ratio is 29. 165 2 and the Correct Process Ratio is 0. 997 1 by combination filtering method. Both are good methods to process image in low illumination.