Research On Wheat Spike Identification Based On Color Features And Improved Adaboost Algorithm

Wheat is a kind of essential crops which has world’s first acreage and production. Inthe world, In the world, a third of the population of wheat as the main food. And, wheatproduction is the cornerstone of the entire agriculture in the national economy but alsooccupies a pivotal position in China. The selection of appropriate varieties of wheat andmaster their characteristics to accurately predict the rate of wheat tiller are of greatsignificance for increasing wheat yields. The traditional method requires a huge amountof labor to pay for breeders. Therefore, to achieve automatic identification of wheat hasimportant practical significance.The traditional methods of wheat production forecast are that: prediction for field ofartificial judgement, satellite prediction, prediction method of the relationship betweennatural conditions and demand analysis and forecasting, the year for forecasting andother methods, these prediction methods all having a certain degree of accuracy but withobvious flaws: to determine the result of human judgment often through experience, andthe difference of more or less personal experience led to predict the existence of bias.satellite prediction use satellite remote image, it has the amount of image capture areaand data for analysis and induction macroscopic, but has a low yield prediction accuracyof wheat in some areas. prediction method of the relationship between natural conditionsand demand analysis based on previous experience, it has a low accuracy and lowreliability.Given the many shortcomings of traditional methods, in order to provide breedersmore reliable, convenient, scientific data, this paper presents the computer multimediatechnology in which wheat production forecast to digital image processing technologybased on combining the color features, pattern recognition, etc, researched the color,shape, texture and other characteristics of wheat and designed of a classifier to identifywheat image count. The main work is as follows:1. Discussed several common color spaces, including space RGB, YUV space, HSVspace. Analysis of the color information in the milky stage of wheat and presented amessage for tone wheat as the main color feature, the use of tonal information byexperiment as wheat color features can largely eliminate the background photo thatexplained wheat color feature has characteristics significantly.2. According to the environmental conditions of wheat fields, designed method anddevice for sample collection. Preprocessed the image and proposed a method of imagesgray-scale based on HSV space H component.3. Introduced the idea of Adaboost algorithm, given a detailed description of theweak classifier training process and the final composition of the cascade classifierprocesses, and Haar features, integral image and other theories. Finally, the experimentalresults show that the use of Adaboost algorithm has a good ability to identify the wheatoccluded. 4. Haar features can only be judged on shape but not texture characteristics. So it ishard to distinguish the wheat blocked, the paper improved Adaboost algorithm, andintegrated textures, shapes of LBP features to replace Haar features, and then combinedwith the color features of wheat to detect wheat. Experiments show that thecharacteristics of the multi-scale LBP more effectively describe the shape and textureinformation, improved algorithms not only improve the detection speed of the algorithm,and it can be detected in wheat cover or inclination, etc, the detection algorithm canimprove the ability to reduce the rate of error detection.