Abstract: Abstract: Color changes of rape leaves are closely related to the variety and quantity of nutrient deficiency. To achieve a fast accurate nutrient diagnose for rape, an intelligent analysis and diagnose method for rape nutrient deficiency was proposed according to color changes of rape leaf caused by nutrient deficiency, which was based on the non-equal quantization of color histogram in the HSV (hue, saturation, value) color space and combined multiple support vector machine (SVM) classifiers. In order to build a validating database, the Hoagland's formula was used to configure nutrient solution and the Kawasaki soilless formula was employed to cultivate a batch of new double-low rapes (the variety "Sun 2009" (Brassica napus L.)). With the soilless culture technique, 5 different growth conditions, i.e. normal, potassium-deficiency, phosphorus-deficiency, nitrogen-deficiency, and borondeficiency, were mimicked to cultivate rape samples. During the cultivation, the pH value of the nutrient solution was in the range of from 6 to 7. First, the nutrient solution was dispensed as base solution and then was diluted with water to make different working liquids for use. To avoid unexpected precipitation caused by chemical reaction, the nutrient solutions were grouped into 3 different base solutions in terms of their chemical characteristics. Next, these base solutions were configured as the original liquids in a certain order and rule. When irrigating, the original liquids were diluted with water in the proportion of 1:10. In an indoor light situation, a Canon EOS600 digital single lens reflex (DSLR) camera, which was equipped with EF-S18-135mm f/3.5-5.6 IS lens and a high-resolution 18 megapixel (MP) APS-C CMOS sensor, was applied to capture the rape leaf images. For each type of nutrient deficiency, 100 labeled leaf images in young seedling were collected and their enclosing rectangle regions were manually cropped to establish the nutrient deficiency database comprising 500 rape leaf images with 5 different types of nutrient deficiency. There were 2 parts in the validation experiment, namely learning classifiers and classification test. In the first part, the active contour segmentation algorithm was first applied to segment the regions of rape leave images in the training dataset. Then the segmented images were converted from the RGB (red, green, blue) color space to the HSV color space. The HSV color histogram of segmented images was extracted and non-equally quantized into a 56-dimentinal feature vector to represent the color differences between different types of nutrient deficiency in rapes. For the constructed training set comprising the feature vectors and labels, 5 SVM classifiers with the RBF (radial basis function) kernel function, were trained by a one-to-many scheme for the recognition of different types of nutrient deficiency. The obtained parameters of the classifiers were saved for the subsequent classification task. In the second stage, the HSV feature of a given rape leaf image in the test set was obtained by the same segmentation and feature extraction procedure as the previous learning stage. Next, the obtained feature vector was input into 5 SVM classifiers in sequence. The class of the classifier whose output was maximal among these classifiers was determined as the final nutrient deficiency class of the input rape leaf image. The experimental results on nutrient deficiency classification indicate that the proposed method can accurately identify the common nutrient deficiency, which provides an effective way for the digitalized and intelligent analysis and diagnose of rape nutrient deficiency.