【摘要】 针对苹果幼果与其背景相似度高的检测问题，图像处理方法被提出来增强目标的注意力机制，并通过YOLOV4目标检测框架进行目标检测任务。近年来，目标检测技术成功应用于人脸识别、自动驾驶等领域，该文针对在苹果幼果检测的近景阶段中，为了提高采集到的数据集中识别效果，提出通过BG方法和RTHTR方法提高幼果目标在图像中的特征值和目标的注意力机制。实验结果表明，经过BG图像处理过程得到的目标检测数据在YOLOV4网络结构中mAP值由YOLOV4网络原数据集的75.38%提升4.14%，经过RTHTR图像处理过程的目标检测数据的mAP值达到82.29%。更多还原

【Abstract】 Aiming at the detection problem of high similarity between the young fruit of apple and its background. Image processing methods were proposed to enhance the target’s attention mechanism,and target detection task was carried out by YOLOV4 target detection framework. In recent years,target detection technology has been successfully applied to face recognition,automatic driving and other fields. In order to improve the recognition effect of collected data in the close-range stage of young fruit detection,the BG method and RTHTR method were proposed to improve the feature value of young fruit in the image and the attention mechanism of the target. The experimental results show that the mAP value of target detection data obtained by BG image processing in YOLOV4 network structure is increased by4.14% from 75.38% of the original data set,and the mAP value of target detection data obtained by RTHTR image processing is up to 82.29%.更多还原