Research on Camera-Based Target Detection Enhancement Method in Complex Environment

A target detection method based on convolutional neural network for multiple complex environments is proposed to address the problem that current target detection methods are prone to miss detection and false detection under complex environment conditions, resulting in reduced detection accuracy. First, an environment recognition architecture is constructed based on convolutional neural networks to classify the images acquired in different environments. Then, the images in different environments are enhanced separately using an image enhancement algorithm. Finally, the enhanced images under different environments are trained based on the YOLOX detection algorithm to achieve real-time detection of the images, and the NUSCENES dataset is used for experimental evaluation. The results show that the proposed detection method improves the average accuracy AP for target detection in sunny, rainy, cloudy and night environments by 6.5%, 13.56%, 6.52% and 6.82%, respectively, improving the detection of small targets.