Detection method for fitting moving target detection frame

The invention discloses a detection method fitting a moving target detection frame. An improved Faster-RCNN deep learning network is constructed, a rotating target detection frame with target motion direction angle information is obtained through regression of target angle information, different angles are set to control extraction of candidate target areas, and video monitoring images are detected. According to the method, target angle information in any direction is introduced into a deep learning target detection task and is improved, so that the target position can be better positioned, and a rotating target detection bounding box which is more fit with the edge of the target is generated to represent the target. 一种贴合运动目标检测框的检测方法，构建改进Faster-RCNN深度学习网络，通过回归目标角度信息得到带有目标运动方向角度信息的旋转目标检测框，设置不同的角度来控制提取候选目标区域，对视频监控图像进行检测。本发明将任意方向的目标角度信息引入深度学习目标检测任务并加以改进，可更好地定位目标位置并生成更加贴合目标边缘的旋转目标检测边界框表征目标。