The Research On Mobile Devices Leaf Recognition Rased On Polar Coordinates

With the advancement of science and technology, smart phones have gradually entered people’s lives. Smartphone application is very large, some applications on image recognition is also emerging in smart phones. However, due to the differences between the phone and PC, the mobile applications need to consider more factors during application development. PC and mobile phone have many differences. Firstly, PC has more computing power than mobile, so the amount of calculation during the mobile application development should be taken into account. Secondly, PC has more memory than the mobile, in the progress of image processing, the phone is likely to cause memory overflow problems if handled properly, so we need to optimize the code to reduce memory usage. Thirdly, phone battery is a scarce resource, excessive amount of computation will consume the phone battery, and phone is more attention to the computational complexity than the PC.Generally, the methods of traditional image feature extraction are based on a Cartesian coordinate system, for example, leaf edges and leaf texture. Although these features are very intuitive, but there are some drawbacks. Firstly, the time complexity of these algorithms is very high. Secondly, the computation of these algorithms is very large. Thirdly, it occupies a lot of memory. For mobile phones, these flaws are fatal.For the features of smartphones and the extensive research of image feature extraction, a feature extraction method based on polar coordinates has been proposed. Leaf images should be converted to a polar coordinate image, then features are extracted in the polar coordinate image, the features include span ratio, surface area ratio, saturation and height change rate. The time complexity of this method is O(n), and it can reduce the amount of memory in the process of extraction because polar images can be stored in a one-dimensional array. Experiments show that the parameters proposed in this study have a stable state, and it has a very good recognition rate in the Flavia leaves set.In addition, this study also develops leaf identification applications based on polar coordinates on Android mobile, and it realizes offline leaf identification and server data synchronization.