Image processing based technique for classification of fish quality after cypermethrine exposure

• Non Destructive image processing technique to identify pesticide treated fish.
• Gill tissue is selected as focal tissue which is segmented for discrimination.
• Support vector machine classification is performed on the extracted features.
• A strategic relationship is established for discriminatory difference of image coefficients.
• The coefficients act as indicative parameter for presence of pesticides.

The quality of fish is primarily dependent on its handling, processing, storage, exposure to contaminants and on climatic variability. Fishes nurtured at fresh and contaminated water exhibit marked differences in quality. Among different contaminants, pesticide is reported as a predominant non-specific menace to fish health and quality. Detection and identification of pesticide residues in fish is a challenging task and requires costly sophisticated instruments. This paper proposes an image processing based non-destructive technique for identifying quality differences between pesticide treated and untreated (control) fish. To evaluate the quality variability, rohu (Labeorohita) fishes were treated with mild dose of cypermethrin for seven days and bio-accumulation status was recorded through GC–MS at post-harvested condition followed by imaging at two days interval. Gill tissue was selected as focal tissue for image processing which was segmented and different features were extracted in wavelet domain using Haar filter. Features were selected up to the third level of decomposition in wavelet domain and analysed for discriminatory features. The discriminatory variations in the different features of images were related to the difference between pesticide treated and untreated fish using strategic image processing techniques. Supervised classification was performed on the extracted features using support vector machine (SVM) classifier. The experimental results indicate that the proposed method is efficient for identification of pesticide treated and untreated fish from the features of the images. The accuracy of identification is high and the computation time is faster enough to make this method efficient as a real time application.

Figure optionsDownload as PowerPoint slide