Recognition of weed seed species by image processing

Recently, in order to increase the speed detection of seeds, the methods based on computer vision is expanded. Since, some characteristics such as colour, morphology and texture show the difference of objects. Therefore, it is necessary that these parameters should be used by machine vision for reorganization of different objects from each other. In this study, identification of four major weed seeds (common vetch, cleavers, cornflower and great bur-parsley), that are widely found in farms of West and North West of Iran, was clone by digital image analysis. Recognizing and removing the weed seeds from the main product is very important. By recognizing the weed seed species from each other, it is possible to determine the percentage of farm pollution to each of weed seeds, and then weed control operations will be applied. The value of products is determined through ratio of weed seeds weight to total weight. Therefore, recognition of weed seed species as the first step in determination of products value is very important. For this purpose, by using a chamber of imaging, some uniform images of samples were acquired. Then, a program was coded in Matlab software for segmentation of the samples. Recognition of weed seeds was based on morphology and colour. For recognizing common vetch, great bur-parsley and cornflower colour features defined in RGB and HSI colour models were used. These colour features were mean (red), mean and variance of saturation component. Recognition of cleavers was done by two morphology features, Shape factor 1 and Shape factor 2. According to the results, total classification accuracy was 98.40%. This shows that the system has great potential to serve as an intelligent recognition system in real applications.