Object detection and recognition of intelligent service robot based on deep learning

Object detection and recognition is the premise and foundation for intelligent service robot to understand the surrounding environment and make intelligent decisions. In this paper, aiming at the accuracy and real-time performance of object detection and recognition of service robot in complex scenes, an end to end object detection and recognition algorithm based on deep learning is proposed. Firstly, the local multi branch deep convolution neural network is adopted to enhance the feature representation capability of the model by enhancing the convolution module function. Then, combining the anchor point mechanism, the object class and position regression prediction is carried out on the multi-layer feature map. When the local features and the global features are fully fused, the natural multi-scale detection and recognition is realized on multiple receptive fields. Finally, a network acceleration module is designed for GPU parallel acceleration on high performance NVIDIA TX1 embedded board. The experiment was carried out on SIASUN second generation intelligent service robot. The experimental results show that the algorithm has both good accuracy and real-time performance.