Inspection methods and devices

Patent metadata

Field	Value
Publication number	US-10013615-B2
Application number	US-201415034021-A
Country	US
Kind code	B2
Filing date	Nov 7, 2014
Priority date	Nov 7, 2013
Publication date	Jul 3, 2018
Grant date	Jul 3, 2018

Abstract

The present disclosure discloses an inspection method and device. The method comprises steps of acquiring a perspective image of an inspected object; processing the perspective image to obtain a region of interest; and automatically detecting the region of interest using a cigarette model, to determine whether the region of interest of the perspective image belongs to a cigarette. In the present disclosure, cigarette detection is implemented on a scanned image of goods, particularly a container, which can avoid the problem of detection vulnerability and poor effect of manual image judgment for the conventional manner, and is of significance in fighting against cigarette smuggling.

First claim

We claim:

1. An inspection method, comprising steps of:
acquiring a perspective image of an inspected object;
processing the perspective image to obtain a region of interest; and
automatically detecting the region of interest using a cigarette model, to determine whether the region of interest of the perspective image belongs to a cigarette;
wherein before the step of automatically detecting the region of interest using a cigarette model, the method further comprises a step of setting up a cigarette model including steps of:
setting up a cigarette image library;
processing images in the cigarette image library to obtain respective regions of interest;
extracting features of the regions of interest of the images in the cigarette image library; and
training a classifier based on the extracted features to generate the cigarette model.

2. The method according to claim 1 , wherein the step of setting up a cigarette image library comprises:
scanning cigarette images of different numbers of cigarettes in various arrangement forms in a container to obtain a positive sample library of the cigarette model; and
collecting images of objects similar to the cigarettes and random objects to form a negative sample library of the cigarette model, wherein no image in the negative sample library comprises a cigarette.

3. The method according to claim 1 , wherein in the step of setting up a cigarette model, scanned images of the cigarettes which are stacked in various rules are divided into multiple modes with different widths, and the positive sample library of the cigarette model is generated based on the multiple modes.

4. The method according to claim 3 , wherein the step of extracting features of the regions of interest of the images in the cigarette image library comprises:
manually marking positions of the cigarettes in units of cases to form a set of positive sample features in the multiple modes; and
randomly extracting multiple samples from the negative sample library for feature extraction to form a set of negative sample features.

5. The method according to claim 1 , wherein the step of training a classifier based on the extracted features comprises:
(1) training the classifier in a case that a set of positive sample features and a set of negative sample features are given, wherein the set of positive sample features are extracted from a plurality of positive samples and the set of negative sample features are extracted from a plurality of negative samples;
(2) classifying positive samples and negative samples using the classifier;
(3) removing negative samples which are easily to be categorized according to a degree of confidence of a result, and re-adding enough random negative samples; and
(4) repeating the above steps (1), (2) and (3), until a change in a classification result of the classifier is small enough or an enough number of iterations is reached.

6. The method according to claim 1 , further comprising: performing a manual markup process or an automatic customs declaration analysis process on an image for which no cigarette is detected, to obtain a new cigarette image and update the cigarette model.

7. The method according to claim 1 , wherein in the step of acquiring a perspective image of an inspected object, the image is scaled to a uniform resolution, and is stretched in gray, so that the whole possible value range is filled with a value domain of the image to implement a normalization operation.

8. The method according to claim 1 , wherein in the step of processing the perspective image to obtain a region of interest, an air part in the perspective image is removed to prevent an air noise from generating a detection result.

9. The method according to claim 8 , wherein in the step of processing the perspective image to obtain a region of interest, a binaryzation process is implemented on the perspective image with an air value using an air brightness threshold method, and a cigarette detection process is implemented only on a part of the image which has brightness lower than the threshold.

10. The method according to claim 1 , wherein the step of automatically detecting the region of interest using a cigarette model comprises:
normalizing a given image to be detected and acquiring a region of interest of the image to generate Histogram of Oriented Gradients (HOG) features of the region of interest;
traversing the HOG features using a sliding window to solve a maximal degree of confidence thereof in each window in multiple modes; and
determining a position where a degree of confidence is larger than a particular threshold as a position where a cigarette is located.

11. The method according to claim 1 , wherein a cigarette is detected in multiple scales.

12. An inspection device, comprising:
a digital radiation photographing apparatus configured to detect an inspected object with a ray to obtain a perspective image of the inspected object; and
an image processing apparatus configured to process the perspective image to obtain a region of interest, and automatically detect the region of interest using a cigarette model, to determine whether the region of interest of the perspective image belongs to a cigarette;
wherein the image processing apparatus is further configured to set up a cigarette model by;
setting up a cigarette image library;
processing images in the cigarette image library to obtain respective regions of interest;
extracting features of the regions of interest of the images in the cigarette image library; and
training a classifier based on the extracted features to generate the cigarette model.

13. The device according to claim 12 , wherein the device is configured to automatically detect cigarette smuggling.

Assignees

• Nuctech Co Ltd

Inventors

• Zhang Li
• Chen Zhiqiang
• Li Qiang
• Zhang Jian
• Gu Jianping
• Cui Jin

Classifications

• G06V20/52

  Surveillance or monitoring of activities, e.g. for recognising suspicious objects (recognising microscopic objects G06V20/69) · CPC title

• G06K9/3241

  Physics · mapped topic

• G06K9/42

  Physics · mapped topic

• G06K9/40

  Physics · mapped topic

• G06K9/38

  Physics · mapped topic