Research On Object Detection Method Based On Multi-source Information Fusion

With the advent of the Internet big data era,object detection has been widely used in many fields such as video surveillance,virtual reality,human-computer interaction and behavior understanding,etc.However,object detection by only single-source information is easily affected by factors such as the device itself,external environment and object diversity,resulting in false detection or missed detection,thus reducing object detection accuracy,while multi-source information has its own advantages and can achieve complementary advantages Therefore,this paper takes the fusion of image information and point cloud information as the entry point to study the object detection method of multi-source information fusion.The details of the research are as follows:(1)A multi-source information fusion object detection model based on Point Fusion Voxel Net(PFVNet)is proposed to address the problem that the object detection model is easily affected by the environment in practical application scenarios.First,based on the point cloud object detection model Voxel Net,the image data processing stream is added,and the point cloud information is fused with the image information using the point-to-point fusion method of data projection and feature indexing,and the feature fusion method of point-to-point feature stitching.Second,data enhancement is performed using stacked voxel feature encoding layers to realize the interaction between the information of each point in the point cloud and the local aggregation information.On this basis,the intermediate convolution layer with regional candidate network is used to aggregate voxel local information and expand the perceptual field to obtain richer shape information.Finally,the proposed PFVNet model is verified to be more effective against the influence brought by the environment and improve the object detection accuracy through comparison experiments and ablation experiments.(2)To address the problems that the object detection model cannot take into account the detection of small object s and the inadequate utilization of features,we introduce the Path-Aggregation Network based on PFVNet and propose a multi-source information multi-scale fusion based on Path-Aggregation Point Fusion Voxel Net(PPFVNet).object detection model based on PPFVNet.First,the semantic information contained in the high-level features is used to obtain the overall object features,and the geometric information contained in the low-level features is used to obtain the small object edge features,and the high-level features and low-level features are fused to enrich the contextual information of the features.Secondly,the complementary fusion of multi-scale features is used to improve the object detection performance.Finally,the proposed PPFVNet model is verified to be effective in improving the performance of object detection through comparison experiments and ablation experiments.