Leveraging smart-phone cameras and image processing techniques to classify mosquito genus and species
Identifying insect species integrates image processing, feature selection, unsupervised clustering, and a support vector machine (SVM) learning algorithm for classification. Results with a total of 101 mosquito specimens spread across nine different vector carrying species demonstrate high accuracy in species identification. When implemented as a smart-phone application, the latency and energy consumption were minimal. The currently manual process of species identification and recording can be sped up, while also minimizing the ensuing cognitive workload of personnel. Citizens at large can use the system in their own homes for self-awareness and share insect identification data with public health agencies.
1. A computerized method of identifying an insect specimen, comprising:
gathering a plurality of digital images of the insect specimen positioned within a respective set of image backgrounds;
extracting image portions from each digital image, wherein the image portions comprise body pixels of image data corresponding to the insect specimen and excluding image background pixels;
converting the body pixels into a selected color space data set;
identifying textural features of the image portions from the selected color space data;
segmenting the image background pixels by detecting edges of the body pixels; and
identifying internal edges of the insect specimen surrounding interior background pixels by:
creating a plurality of Gaussian models for the image background pixels that are external to the insect specimen;
determining at least one probability that an intensity value for a subject pixel is greater than a selected threshold calculated from the image background pixels;
tagging the subject pixel as background; and
segmenting the subject pixel.
2. The computerized method of claim 1 , wherein extracting image portions comprises segmenting the image background pixels in the digital images.
3. The computerized method of claim 1 , further comprising tracking contours of the insect specimen by saving body pixels corresponding to the edges.
4. The computerized method of claim 1 , further comprising detecting the edges by comparing changes in intensity values across the body pixels and the image background pixels.
5. The computerized method of claim 4 , wherein comparing changes comprises calculating intensity derivative values across the digital images and comparing changes in the derivative values between at least one body pixel and at least one image background pixel.
6. The computerized method of claim 5 , further comprising identifying contours on the insect specimen by identifying and saving intensity values and direction of intensity value derivatives for the edge pixels.
7. The computerized method of claim 6 , further comprising comparing sets of pixels corresponding to contours with the digital images to confirm the image background pixels.
8. A computerized method according to claim 1 , further comprising identifying the textural features based upon local binary patterns established for each body pixel.
9. A computerized method according to claim 1 , further comprising calculating Haralick textural features for the body pixels.
10. A computerized method according to claim 1 , further comprising:
identifying first texture data sets based upon local binary patterns established for each body pixel;
calculating Haralick textural feature data sets for the body pixels;
fitting the first texture data sets and the Haralick textural feature data sets into a sub-space that is of a smaller dimension than the body pixels.
11. A computerized method according to claim 10 , further comprising identifying a linear combination of sub-space features, wherein the respective features exhibit minimal correlation.
12. A computerized method according to claim 11 , further comprising clustering test features from the body pixels and matching clustered test features with the sub-space features.
13. A computerized method of identifying an insect specimen, comprising:
gathering a plurality of digital images of the insect specimen positioned within a respective set of image backgrounds;
extracting image portions from each digital image, wherein the image portions comprise body pixels of image data corresponding to the insect specimen and excluding image background pixels;
converting the body pixels into a selected color space data set;
identifying textural features of the image portions from the selected color space data set;
identifying first texture data sets based upon local binary patterns established for each body pixel;
calculating Haralick textural feature data sets for the body pixels; and
fitting the first texture data sets and the Haralick textural feature data sets into a sub- space that is of a smaller dimension than the body pixels.
14. A computerized method according to claim 13 , further comprising identifying a linear combination of sub-space features, wherein the respective features exhibit minimal correlation.
15. A computerized method according to claim 14 , further comprising clustering test features from the body pixels and matching clustered test features with the sub-space features.