Pedestrian detection method using ReCNN fused context information

The invention provides a pedestrian detection method using ReCNN (Recurrent Convolutional Neural Network) fused context information. The method comprises the following steps: step A, extracting features including context information in an input image; step B, predicting a pedestrian position in the input image by using an ReCNN, and acquiring a mask map for predicting the pedestrian position; stepC, determining the position of a pedestrian in the input image; and step D, training parameters of the ReCNN fused with the context information. The pedestrian detection method using ReCNN fused context information fuses the image information of multiple continuous frames of a video by using the ReCNN, and can obtain a more accurate mask map over a method using single image information, thus improving the accuracy of pedestrian position detection. 本发明提供了种利用ReCNN融合上下文信息的行人检测方法，包括以下步骤：步骤A：提取输入图像中包含上下文信息的特征；步骤B：利用ReCNN对输入图像中的行人位置进行预测，获取预测行人位置的掩码图；步骤C：确定输入图像中行人的位置；步骤D：训练融合上下文信息的ReCNN网络的参数；本发明所述的种利用ReCNN融合上下文信息的行人检测方法利用Re