Fast detection algorithm for golden crane fly based on region growing and SVM

A fast detection algorithm for golden crane fly based on the traditional image processing methods and Support Vector Machine was designed to detect the golden crane fly in real environment faster. In this scheme，pest was realized in the process of segmentation which simplified the image processing. Besides，the SVM method not only supported small sample data training of classifier，but also it reduced the number of samples in the training. We took 100 pictures of golden crane fly in the natural environment as the materials for the classifier training and got high recognition rate of the classifier. Meanwhile，we also realized the algorithm using the classifier combined with the traditional image processing methods. By detecting 80 pictures of golden crane fly， it showed that the recognition correct rate for golden crane fly reached more than 90%. Besides，for a clear target in our picture，the detection time of algorithm was less than 0.2 second. For the faster running speed and higher precision of the algorithm，it could provide the technical support for fast monitoring of pests on the field and had good application prospect.