Research On Target Detection Algorithm For Outdoor Community Static Scene

At present,there are significant challenges in detecting and recognizing targets in outdoor scenes when designing navigation systems and auxiliary tools for visually impaired individuals.Researching an effective outdoor scene object detection algorithm can greatly improve the safety and convenience of travel for visually impaired individuals.However,the image data collected in outdoor scenes usually includes low light images,which have problems such as low brightness and loss of detail information,posing new challenges to object detection algorithms.This article focuses on the difficulty of object detection in outdoor community scenes,conducts research on low light image processing methods and object detection algorithms,and designs an object detection system in outdoor community scenes,providing theoretical support for automatic navigation of visually impaired outdoor communities.1.Propose an improved low illumination image enhancement method.Based on the Retinex-Net network,improve it.First,the input image is converted from RGB domain to HSV color space,and the reflection image of V component is introduced into denoising Convolutional neural network for denoising.Then,the illumination image of V component is enhanced through attention mechanism.Finally,the components are fused and converted back into RGB space to obtain the enhanced image.Experiments have shown that the low light image enhanced using the algorithm proposed in this paper has improved brightness,prominent details,small image distortion,and is realistic and natural.It is superior to other algorithms in terms of subjective perception and objective evaluation indicators.2.An improved outdoor object detection algorithm based on YOLOv4 is proposed.Firstly,the Focus module is inserted into the backbone network.Secondly,for the extraction of image feature details,a spatial serrated hole convolution structure is used to replace the original network structure for strengthening,and the neck network structure is cropped to reduce the network weight.Finally,the bidirectional feature pyramid structure is used to enhance the model’s ability in shallow prediction and deep localization.The experiment shows that on the outdoor scene dataset constructed in this article,the improved algorithm significantly improves detection accuracy and speed compared to the original YOLOv4 algorithm.3.Build an outdoor community scene object detection system.The improved low light image enhancement algorithm and the improved enhanced object detection algorithm are fused to form a static outdoor community scene object detection system.The effectiveness of the algorithm in this paper is verified through testing the system in actual outdoor community scenes.The comparison of object detection performance with images without image enhancement proves that the system has better object detection performance and effectiveness.