Vehicle dense target detection method based on deep learning

Abstract

The invention discloses a vehicle dense target detection method based on deep learning. The method comprises the following steps: S1, constructing and training a multi-scale dimensionality reduction convolution feature extraction network; S2, extracting a multi-scale dimensionality reduction feature map of the to-be-detected image through a multi-scale dimensionality reduction convolution feature extraction network; S3, generating a priori knowledge anchor box based on the historical image; S4, generating all target candidate areas in the to-be-detected image; And S5, carrying out ROIpooling processing on the target candidate area to obtain a detection result with a vehicle dense target. According to the invention, the characteristic of large scale difference of the dense target is considered, and on the basis of the faster-rcnn network, the thought of multi-stage multi-resolution and multi-size dimensionality reduction convolution feature extraction and shape prior-based anchor window generation is provided, so that the detection capability of a multi-scale dense model is improved, the problem of information loss in the existing related detection method is effectively solved, and automatic identification and discrimination of dense targets are realized.

Description

technical field [0001] The invention belongs to the technical field of intelligent transportation, and in particular relates to a vehicle dense target detection method based on deep learning. Background technique [0002] With the acceleration of our country's modernization process, the construction speed of urban infrastructure cannot meet the needs of rapid economic growth at all, and crowding of people and vehicles often occurs, resulting in traffic safety, road congestion, environmental pollution, etc. The problem is becoming more and more prominent. One of the main reasons for frequent accidents is the large traffic flow on the road, crowded vehicles, and illegal mixed traffic, which are directly related to the density of vehicles. The densely distributed vehicles in the actual scene will not only directly lead to the occurrence of traffic accidents, but also affect the service capabilities of urban infrastructure. If the intelligent traffic analysis system can detect...

AI Technical Summary

Problems solved by technology

[0004] In view of the above-mentioned deficiencies in the prior art, the vehicle dense target detection method based on deep learning provided by the present invention solves the problem that the existing related target detection method does not consider the target characteristics, so that the target cannot be recognized normally and cannot detect all complex scenes. target problem

Images