In order to detect fertilized eggs nondestructively to improve hatching rate, this paper uses the method of image processing and Learning Vector Quantization neural network to identify fertilized eggs. Firstly, we use image collection device to collect images of the unfertilized and fertilized eggs and extract the feature of egg image, and then determine 5 principal component characteristics of the egg shape. Learning vector quantization neural networkis 5 dimensional input and 1 dimensional outputs.Finally,we use Genetic Algorithm to optimize the weights and threshold of neural network, which can be used to predict the condition of fertilization. The experiment shows that, compared with the traditional LVQ neural network, it is more accurate to recognize the fertilized eggs when using optimized LVQ neural network by genetic algorithm. The rate can reach 98.21%, which meets therequirements of recognizing fertilized eggs.
Genetic algorithm with deterministic crossover for vector quantization
