Standard methods for inexpensive pollen loads authentication by means of computer vision and machine learning

1. Introduction 2. Review on computer vision and classification methods for pollen recognition 3. General overview of the standard method and hardware requirement 4. Description of the computational techniques of the methodology 4.1 Segmentation algorithm 4.2 Homogenizing the pollen loads of the image by mean shift filtering 4.3 One-class multi-classification algorithm based on distances 5. Validation results and discussion 5.1 Experimental data and image processing results 5.2 Numerical performance of the multi-classification algorithms 5.3 Validation of the complete prototype 6. Conclusions 7. References Summary We present a complete methodology for authenticating local bee pollen against fraudulent samples using image processing and machine learning techniques. The proposed standard methods do not need expensive equipment such as advanced microscopes and can be used for a preliminary fast rejection of unknown pollen types. The system is able to rapidly reject the non-local pollen samples with inexpensive hardware and without the need to send the product to the laboratory. Methods are based on the color properties of bee pollen loads images and the use of one-class classifiers which are appropriate to reject unknown pollen samples 2 when there is limited data about them. The validation of the method is carried out by authenticating Spanish bee pollen types. Experimentation shows that the proposed methods can obtain an overall authentication accuracy of 94%. We finally illustrate the user interaction with the software in some practical cases by showing the developed application prototype. Short title Automatic pollen loads authentication