Authors: Haidi Yuan; Wei Li
Addresses: Foundation Experiment Teaching Center, An Hui Sanlian University, Hefei, 230601, China ' Information Center, Anhui Vocational College of Press and Publishing, Hefei, 230601, China
Abstract: Aiming at the complex mine environment, a multi-target detection method based on machine learning is proposed. Firstly, aiming at the complex mine environment, a multi-target detection framework based on fast RCNN is established. Then, through the softmax activation function of the regional suggestion strategy, the probability that each feature point in the feature mapping image belongs to the foreground is solved. Finally, the feature pyramid network (FPN) and multitask loss are added to the fast RCNN network to improve it. Experiments show that this method is accurate and fast in selecting the recommended boundary frame of the image area. The accuracy of multi-target detection for equipment, personnel and so on is more than 91%, and the detection results are less affected by noise.
Keywords: machine learning; complex mine environment; multi-target detection; faster RCNN; feature extraction; region proposals.
International Journal of Reasoning-based Intelligent Systems, 2022 Vol.14 No.4, pp.201 - 207
Received: 15 Apr 2022
Accepted: 25 May 2022
Published online: 31 Oct 2022 *