A passive millimeter wave image human body target segmentation method for security inspection of prohibited objects

Abstract

The invention relates to a passive millimeter wave image human body target segmentation method for security inspection of prohibited objects, which utilizes a passive millimeter wave imager to collectthe passive millimeter wave image of the human body target, and directly obtains the segmentation result of the human body target region by means of an end-to-end depth semantic convolution neural network. Because the whole DNN network adopts the end-to-end design idea, it does not need to pre-process and post-process the image, so the use process is very simple. In the improvement of the segmentation precision, the U-Net basic network structure is used for reference. The symmetric deep residual neural network Resnet50 is used in the coding segment and the decoding end of the network respectively. Because Resnet50 has better feature extraction cability than the VGG16 backbone network. so the segmentation accuracy of the DNN network model designed by the present invention is further improved compared to the classic U-Net network.

Description

technical field [0001] The invention relates to a method for segmenting a human object in a passive millimeter wave image based on a deep convolutional neural network for security inspection of contraband, and belongs to the field of security technology. Background technique [0002] In recent years, terrorist attacks and violent crimes around the world have been constantly changing and escalating, posing a great threat to public security. In addition, the inevitable huge traffic passenger pressure in megacities also puts enormous pressure on security work. Therefore, it is of great significance to study the detection technology of hidden and prohibited articles in large venues and public security checkpoints of transportation hubs. [0003] At present, security equipment widely used in public places mainly includes: metal detection gates, hand-held metal detectors, and X-ray security detectors. At present, the metal detectors commonly used in public security ports can onl...

AI Technical Summary

Problems solved by technology

[0013] The defect of U-Net is that the main structure of its encoding and decoding segment network is VGG16 full convolutional neural network, so the feature extraction ability is relatively limited.

Images