Pose Detection of Object on Transmission Belt Based on R-SSD

The traditional object detection algorithm can only obtain the horizontal detection frame. In this paper, to solve the problem of large background noise and missing angle information in the detection frame, a bottle target detection model based on SSD (Single Shot multi-frame Detector) algorithm is proposed. The convolutional neural network is used as the feature extraction unit, which realized the gradient bounding box regression by introducing angle prediction. In order to improve the accuracy of the model, the feature pyramid structure is introduced into the feature extraction network, feature pyramid structure enables deep semantic information in the network to be combined with shallow layers. The prior box is determined by K-means clustering algorithm, it is used to make the network model more suitable for the bottle target. The experimental results show that the proposed algorithm is better than the original SSD algorithm, the AP value reaches 90.7%, and the real-time performance of the algorithm is also greatly improved. In summary, the position + size + angle simultaneous detection is more suitable for the target pose recognition application, such as grabbing the bottle in garbage sorting.