Wheat Spike Recognition Technology Based On Color And Texture Features

Wheat is an important food crop in China,and there are many methods for predicting its yield.The number of panicles per unit area is an important component of the estimated yield.The research object is mature wheat.At this time,wheat is close to harvest,which is the final stage of the whole growth cycle.Affected by other(weather,pest)small probability of reducing production conditions,the wheat field image is automatically counted at this time,and then the yield evaluation is close to the harvest yield,which is more accurate and faster than manual evaluation.In view of the fact that there are some weeds in the background of wheat field under natural conditions,the yellowing of wheat ears and the land of stems and leaves are not easy to distinguish from the color point of view.How to effectively isolate the background and identify wheat has become the focus of this paper.In order to solve the background noise of wheat photographing images,this paper takes 30 images of 200 wheat fields monitored and collected by Shangqiu Academy of Agricultural Sciences at the end of May 2018 as the research object,and preprocesses the collected images to enhance the images.The feature separates the weeds,and then uses the texture features to identify the wheat ears,and finally counts the wheat ears based on the corner points.For the weeds in field image,the h-pass value based on hsv color space was adopted,and the threshold method was used to segment the segmentation effect.The weed segmentation accuracy was 97.8%.The texture feature was used to remove the weed image.Spike recognition,gridding the image into small pixel blocks,extracting four texture eigenvalues based on gray level co-occurrence matrix for different categories(wheat,stem,leaf,land),and using the improved k-means algorithm to aggregate texture feature values The wheat ear was identified by morphological processing.The wheat ear skeleton was extracted by morphological treatment.The number of artificial markers was compared and the results were analyzed based on Harris corner detection.Finally,the accuracy of wheat ear identification counting is 94.65%.The research method of this paper can provide ideas for the identification and segmentation of other crops,and provide support for crop remote information management system.The research conclusions can provide guidance and basis for wheat estimation in complex field environment.