Hoh Xil Tibetan antelope detection method based on convolutional neural network

Abstract

The present invention discloses a Hoh Xil Tibetan antelope detection method based on a convolutional neural network. The objective of the invention is mainly to solve the problems that the Tibetan antelope population statistics is slow in speed and not accurate in the prior art. The method comprises the steps of: 1) constructing a training data set, performing marking of Tibetan antelope images obtained through enhancement of the collected data, and obtaining a training data set; 2) employing a k-means algorithm to perform target frame dimension clustering of the training data set, obtaining the most effective priori frame information of the Tibetan antelope data set, and constructing a convolutional neural network with 19 convolutional layers and 5 maximum pooling layers; 3) performing training of the Tibetan antelope data by employing the convolutional neural network to obtain a Tibetan antelope detection and classification model; and 4) employing a trained model to perform detectionof the Tibetan antelope images, draw concrete positions of Tibetan antelopes and mark classes of the Tibetan antelopes. The Hoh Xil Tibetan antelope detection method based on a convolutional neural network can more rapidly and accurately perform detection and classification of the Tibetan antelopes, and can be used for scientific protection and management of the Tibetan antelopes in natural reserves.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and further relates to a method for detecting the Hoh Xil Tibetan antelope based on a convolutional neural network. It can be used for the statistics of Tibetan antelope targets in Hoh Xil Nature Reserve and the classification of male and female Tibetan antelope individuals, which is convenient for reserve management agencies to grasp the dynamics and trends of Tibetan antelopes, thereby providing a basis for scientific management of Tibetan antelope populations. Background technique [0002] Tibetan antelope, as a first-class national protected animal, is included in Appendix I of the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES), which is mainly distributed in alpine desert meadow grasslands and plateau grasslands in Qinghai, Tibet, Xinjiang and other places in China etc. environment. The unique ecological environment makes the protection of Ti...

AI Technical Summary

Problems solved by technology

[0003] At present, the research on the population statistics and distribution of Tibetan antelope is still in the stage of manual operation

For example, "Number and distribution of Tibetan antelope, Tibetan wild ass and wild yak in the Altun Mountain Nature Reserve" published by Lu Feiying and others from the School of Life Sciences of Capital Normal University (Journal of Beijing Normal University (Natural Science Edition), 2015 (4) :374-381.) used the line transect survey method to investigate the quantity and distribution of three ungulates, Tibetan antelope, Tibetan wild donkey and wild yak. ​​The disadvantage of this method is that the survey period is long, and the statistical population is An approximate number cannot accurately count the distribution of the number of Tibetan antelopes, etc., and it cannot distinguish between male and female Tibetan antelopes in complex situations.

Images