Machine part sorting is important and monotonous in smart factory. In this paper, an auto-sorting system is proposed based on the deep learning method. In the proposed system, an industrial objection detection network combined with a robotic arm controlling system is designed to automatically and efficiently complete machine part sorting. Region-based full convolutional network (R-FCN) is applied for locating and recognizing different types of images of industrial object models. After comparison and simulation analysis, it illustrated that the R-FCN model trained with enough labeled data can efficiently and accurately recognize the object from the images captured by visual sensors. Furthermore, with enough data, the network can be robust to view angle rotation both vertically and horizontally, and a small part of overlapping of object will not mislead the judgment of the network in most situations. The case study results illustrate that the position and type of objects can be successfully detected. The code will be available publicly at https://github.com/tianwangbuaa/.

A reliable prediction of traffic participants’ trajectories is challenging for automated driving. We attempt to integrate human trajectory’s temporal and spatial reciprocal consistency into the prediction network structure. Moreover, humans are intention-driven agents, and modelling human intentions will continuously provide more accurate and detailed information for future trajectory prediction. In this paper, we propose a Reciprocal Consistency Prediction Network (RCPNet), a bi-directional prediction network consisting of the forward and the backward with a similar structure. The prior work of human trajectory prediction is mainly a unidirectional process, while RCPNet fully uses the past and future trajectory information at reverse temporal scales. By training the backward part to optimize the whole network parameters based on the reciprocal consistency, we can improve the prediction accuracy of the forward network. In particular, the framework incorporates a destination variational auto-encoder (DVAE) to estimate the target intention of humans and generate multi-modal trajectory prediction. In order to adapt to the offline prediction situation, we refer to the calculation method of correlation between two variables to present “Predictive Confidence” to clarify the reliability of offline predicted trajectories. Experimental results show that the RCPNet improves state-of-the-art performance on the SDD and ETH/UCY benchmarks.