Study On Wheat Spike Recognition Based On Image Processing

Wheat is an important food crop and its acreage and the total of wheat are the first in the world. The wheat is regarded as the main food by more than one third of the word’s population. Wheat production has an important role in the national economy, and it is the foundation of the entire agricultural production. The choice of wheat varieties is the key to the high yield of wheat. The yield prediction of new varieties and the automatic calculation of the tillering rate are both constraints of breeders’heavy labor. Therefore, the recognition research of wheat image has important scientific significance and value.At present, there are few kinds of wheat yield prediction methods, such as forecasting in the field by artificial judgment, the capacitance forecasting production, remote sensing production forecast, the analysis forecasting method by the relationship between climate and supply and demand, the year’s harvest for forecast, and so on. However, all of the above mentioned methods have some shortcomings. Personal experience judement would lead to different predictable result because of Individual differences. Measuring wheat density and tillering rate are difficult. Using capacitance method production forecast, and it cost much too high also. Density and low accuracy are the drawbacks though capturing images by satellite, and it it difficult to achieving small-scale breeding and estimates. Forecasting by the method of climate and the relationship between supply and demand, forecasting is based on the year’s harvest can both do a wide range of overall rough estimate. Wheat image recognition and analysis could effectively solve these problems which based on computer image processing and pattern recognition technology. These method could provide the basis for breeding species tesing and the scientific prediction of yield. The main purpose of this paper is to research the wheat spike identification mainly by computer image processing technology. As well this paper involves in other knowledge such as photograph, biological characteristics, pattern recognition etc. The main contents of this paper are as follows:1. The image of wheat spike collectionThe method of shooting image is researched on to advance the identification accuracy by computer and reduce restrictive conditions when collect image. In order to reduce background interference, adjusting the camera imaging modalities and shooting the panicle separately in the field ridge gap.2. The image of wheat spike pre-processImage grayscale and binary: First, the spike images were processed by traditional grayscale method to test. After compared the processed image, a method was found out in research that wheat spike image gray-binary processing using improved G-component method. Image enhancement: After binarization processing, there were isolated noise points in the image. By consulting datum, fast parallel median filter was adopted to smoothing image.Background Segmentation: Sobel operator edge detection method was adopted in this paper for partitioning background. Finally, in order to make the target image more specific, binary closing operation method of the morphology was filled the hole in the split graph of wheat spike.3. The characteristic of wheat spikeShape features, texture features of spike were extracted in this paper. There were 7 characteristic parameters of Hu moments in shape feature. And 4 characteristic values of texture feature were picked up such as second moment, moment of inertia, inverse difference moment and grayscale correlation.4. The wheat spike identificationUsing the neural network to identify the wheat spike, the BP neural network was constructed in accordance with the shape characteristic parameters and the feature values. Then the network weight was adjusted by training samples. Spike recognition rate was 92%.The innovative aspects of this paperA quick and simple improvement component of image gray method was proposed. This method could make the green spike effectively separated from the khaki background mixing small amount of weed.