Research Of Plant Nutrient Deficiency Based On Digital Image

In recent years, with the development of digital image processing and pattern recognition technology, its application in agriculture has become increasingly widespread. In the diagnosis of plant nutrition, there are many deficiencies in traditional diagnostic methods; we use the digital diagnostic techniques for diagnosis, with fast, efficient and practical features, which has important theoretical significance and practical value in this area. In this paper, we study the type of recognition for plant nutrient deficiency, including the image acquisition, image preprocessing, image segmentation, image feature extraction, pattern recognition.For noise images, through comparing frequency domain filtering with spatial filtering, we finally proposed median filtering to preprocess. Specifically, we sorted the pixel of filter sliding window, and the final output was the mid-value of the sequence, which can not only remove the noise but also protect the edges of the picture.In order to improve the accuracy and effectiveness of segmentation for deficiency image, we proposed a segmentation method based on pulse coupled neural network. Firstly, we selected the processing object in the RGB and HSV space, using the principle of maximum entropy mechanism of entropy. Then we used the pulse coupled neural network(PCNN) model for lesion binary segmentation. Contrasted with the traditional OSTU, we can conclude that that the effect of PCNN is better segmentation process, robustness, high accuracy segmentation.In pattern recognition, we proposed support vector machine applying to plant nutrient deficiency diagnose. Firstly, we determined the feature value for support vector machine classification. We chose the RGB and HSV color space as color features, and chose the mean and variance of energy, entropy, contrast, correlation as texture features. Secondly, the support vector machine was applied to classification pattern recognition. We selected a different combination of features to classify.Through genetic algorithm to optimize the parameters of support vector machine, you can see the final classification accuracy improved, which plays an optimization effect.