基于图像识别技术的夏玉米生育期识别方法初探

通过对河南省鹤壁市农业气象试验站夏玉米整个生育期的田间观测,利用＂农情1号＂田间监测设备定期采集不同生育期玉米的图像资料,分析测点内玉米群体图像特征,结合多种颜色特征参数实现对田间玉米主要生育期的识别。针对玉米不同生育期,分别采用RGB颜色空间提取绿色像素值占整幅图像比例和HSL颜色空间提取黄色像素值占整幅图像比例进行判定;针对田间杂草影响图像颜色的提取,提出基于区域标记的小面积消去算法消除图像中部分杂草的干扰;针对图像中湿润土壤对提取黄色像素值的干扰,提出结合HSL颜色空间S分量消除湿土颜色。经过与人工观测对比,结果表明：利用颜色特征参数对夏玉米主要生育期进行识别是可行的,正确识别率达到94.26%。

Initial Exploration of Maize Phenological Stage Based on Image Recognition

Through the observation and research of whole phenological stage of summer maize in agro-meteorological station in Hebi city,Henan province and by means of field monitoring image capture device to collect different maize image data and analyzing image feature of group images as well as parameters with a variety of colors,the identification of the main phenological stage can be attained.For different phenological stage,RGB color space to extract green pixel value and HSL color space to extract yellow pixel value are used to determine the proportion of developmental period.For the weed effect on the extraction image color,proposed using small area elimination algorithm based on region labeling to eliminate weed impact in pictures.For wet soil on the extraction of the interference of the yellow pixel value in pictures,proposed combining S value of HSL color space to eliminate the color of wet soil.After comparison with the manual observation,the results showed that it is available to identify maize developmental period by using color parameters.The correct recognition rate reaches 94.26%.