Special scene desensitization gelatinization method for automobile data recorder

Abstract

The invention discloses an automobile data recorder special scene desensitization gelatinization method, which is characterized in that a sensitive image detection identification model is constructed based on a feature detection model in an OpenCV SDK, and the method comprises the following steps: loading a dlib library and a license plate detection model in the feature detection model in the OpenCV SDK; obtaining a license plate sample set, a uniform sample set and a vehicle sample set; creating a positive sample file and a negative sample file; training a sensitive image detection and recognition model through the positive sample description file and the negative sample description file; and identifying and judging video data in the automobile data recorder through the trained sensitive image detection and identification model, and if the video data are identified and judged as sensitive images, performing gelatinization by adopting a fuzzy algorithm. According to the video image feature detection model based on the OpenCV SDK, privacy protection of special scenes and special crowds is improved, and the privacy leakage problem of the special scenes and special information is effectively solved through intelligent recognition and accurate fuzzy processing.

Description

Technical Field

[0001] The invention belongs to the technical field of driving recorders, and in particular relates to a desensitization and gelling method for special scenes of driving recorders. Background Art

[0002] With the rapid development of science and technology, video surveillance and image recognition technologies have been widely used in various fields, improving the efficiency of social governance and public services. However, this also brings with it the risk of privacy leakage, and privacy protection is particularly important. At present, the processing of sensitive information mainly relies on physical blocking or watermarking, which are mostly based on fixed fuzzy rules and cannot be adaptively adjusted according to the particularity of the scene and information. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a desensitization and fuzzification method for special scenes of a driving recorder in view of the deficiencies of the above-mentioned prior art, which can fuzzy the identified sensitive information.

[0004] The technical solution adopted by the present invention is: a method for desensitizing and blurring a special scene of a driving recorder, comprising the following steps:

[0005] (1) Building a sensitive image detection and recognition model based on the feature detection model in the OpenCV SDK, including loading the dlib library and license plate detection model into the feature detection model in the OpenCV SDK;

[0006] (2) Obtaining a license plate sample set consisting of special license plate image data and ordinary license plate image data, obtaining a uniform sample set consisting of public servant uniform image data and daily clothing image data, and obtaining a vehicle sample set consisting of special vehicle image data and ordinary vehicle image data;

[0007] (3) firstly, the image data in all sample sets are adjusted to a uniform size, and then the image data is converted into a binary image;

[0008] (4) Using the special license plate image data, public servant uniform image data and special vehicle image data as positive sample images, specifying the folder of the positive sample images, the size of the positive sample images and the name parameters of the positive sample output description file, and then generating a positive sample description file;

[0009] (5) Using ordinary license plate image data, daily clothing image data, and ordinary vehicle image data as negative sample images, specifying the negative sample image folder, negative sample image size, and negative sample output description file name parameters to generate a negative sample description file;

[0010] (6) Training the sensitive image detection and recognition model by setting training parameters and training output results through the positive sample description file and the negative sample description file, and obtaining the trained sensitive image detection and recognition model after the training is completed;

[0011] (7) Obtain the video file in the driving recorder, and use the capread function to read each frame in the video file, and convert the read video frame into a binary image. The converted binary image is recognized by the trained sensitive image detection and recognition model. If it is recognized as a sensitive image, the fuzzy algorithm is used to blur the sensitive image.

[0012] Preferably, the license plate detection model is loaded via the CascadeClassifier class in the OpenCV library.

[0013] The beneficial effects of the present invention are as follows: the present invention improves the privacy protection of special scenes and special groups based on the video image feature detection model of OpenCV SDK, and effectively solves the privacy leakage problem of special scenes and special information through intelligent recognition and precise blur processing. It has the characteristics and advantages of intelligent recognition, precise blur, real-time feedback, adaptive optimization and wide applicability. DETAILED DESCRIPTION

[0014] The present invention will be further described in detail below with reference to specific embodiments.

[0015] Example

[0016] The method for desensitizing and blurring a driving recorder in a special scene provided in this embodiment includes the following steps:

[0017] (1) Building a sensitive image detection and recognition model based on the feature detection model in the OpenCV SDK, including loading the dlib library and license plate detection model into the feature detection model in the OpenCV SDK;

[0018] (2) Obtaining a license plate sample set consisting of special license plate image data and ordinary license plate image data, obtaining a uniform sample set consisting of public servant uniform image data and daily clothing image data, and obtaining a vehicle sample set consisting of special vehicle image data and ordinary vehicle image data;

[0019] (3) firstly, the image data in all sample sets are adjusted to a uniform size, and then the image data is converted into a binary image;

[0020] (4) Using the special license plate image data, public servant uniform image data and special vehicle image data as positive sample images, specifying the folder of the positive sample images, the size of the positive sample images and the name parameters of the positive sample output description file, and then generating a positive sample description file;

[0021] (5) Using ordinary license plate image data, daily clothing image data, and ordinary vehicle image data as negative sample images, specifying the negative sample image folder, negative sample image size, and negative sample output description file name parameters to generate a negative sample description file;

[0022] (6) Training the sensitive image detection and recognition model by setting training parameters and training output results through the positive sample description file and the negative sample description file, and obtaining the trained sensitive image detection and recognition model after the training is completed;

[0023] (7) Obtain the video file from the dashcam, and use the capread function to read each frame in the video file, convert the read video frame into a binary image, and use the detectMultiScale function to detect the license plate area to identify special vehicles and special license plates. The license plate detection model is loaded through the CascadeClassifier class in the OpenCV library; the feature detector is loaded using the dlib library and applied to the video frame to identify the special clothing of office workers; the converted binary image is recognized by the trained sensitive image detection and recognition model. If it is recognized as a sensitive image, the fuzzy algorithm is used to blur the sensitive image.

[0024] The above description is only a preferred implementation manner of the present invention, but the protection scope of the present invention is not limited thereto, and any modification and replacement based on the technical solution and inventive concept provided by the present invention should be included in the protection scope of the present invention.

Claims

1. A method for desensitizing and pasting special scenes of a driving recorder, characterized in that: The following steps are involved: (1) Building a sensitive image detection and recognition model based on the feature detection model in OpenCV SDK, including loading the dlib library and license plate detection model into the feature detection model in OpenCVSDK; (2) Obtaining a license plate sample set consisting of special license plate image data and ordinary license plate image data, obtaining a uniform sample set consisting of public servant uniform image data and daily clothing image data, and obtaining a vehicle sample set consisting of special vehicle image data and ordinary vehicle image data; (3) firstly, the image data in all acquired sample sets are adjusted to a uniform size, and then the image data are converted into a binary image; (4) Using the special license plate image data, public servant uniform image data and special vehicle image data as positive sample images, specifying the folder of the positive sample images, the size of the positive sample images and the name parameters of the positive sample output description file, and then generating a positive sample description file; (5) Using ordinary license plate image data, daily clothing image data, and ordinary vehicle image data as negative sample images, specifying the negative sample image folder, negative sample image size, and negative sample output description file name parameters to generate a negative sample description file; (6) Training the sensitive image detection and recognition model by setting training parameters and training output results through the positive sample description file and the negative sample description file, and obtaining the trained sensitive image detection and recognition model after the training is completed; (7) Obtain the video file in the driving recorder, and use the capread function to read each frame in the video file, and convert the read video frame into a binary image. The converted binary image is recognized by the trained sensitive image detection and recognition model. If it is recognized as a sensitive image, the fuzzy algorithm is used to blur the sensitive image.

2. The method for desensitizing and pasting special scenes of a driving recorder according to claim 1, characterized in that: The license plate detection model is loaded through the CascadeClassifier class in the OpenCV library.