Application of SSD framework model in detection of logs end
Hao Tang, Kejian Wang, GU Jian-cai, Xiaoye Li, Wenhao Jian · Journal of Physics Conference Series · 2020
Abstract External factors in the natural environment interfere with the detection of logs end. A method for detecting the logsend using an SSD model in a natural scene is proposed. The use of default frame in SSD can realize the extraction and utilization of target features with different scales and improve the detection accuracy. By manually marking the foreground area of the logs end, the interference of the features of the background area is reduced, and the convolution learning of the feature of the target area is enhanced. The performance evaluation and comparison test of the method showed that the accuracy rate was 94.87% and the recall rate was 91.34%. Compared with the traditional method, it reduces the occlusion effect caused by light and shooting angle, and improves the detection accuracy.