CN115830288A - Pinhole camera shooting intelligent terminal detection technology based on ToF imaging - Google Patents

Abstract

The invention provides a technology for detecting a pinhole camera shooting intelligent terminal based on ToF imaging, and relates to the technical field of pinhole camera shooting detection. This pinhole intelligent terminal detecting system that makes a video recording based on ToF formation of image includes: intelligent detection terminal, it includes: the system comprises a ToF shooting module and a ToF image processing module, wherein the ToF shooting module collects images and sends the images to the ToF image processing module to operate a target detection algorithm for detection. The invention provides a pinhole camera shooting intelligent terminal detection technology based on ToF imaging, which adopts active infrared type ToF to collect optical images, is less influenced by natural light, has high imaging precision, adopts a target detection model based on a depth network to analyze and judge the ToF images, and has strong user friendliness and low omission factor and false detection rate.

Description

Pinhole camera shooting intelligent terminal detection technology based on ToF imaging

Technical Field

The invention relates to the technical field of pinhole camera shooting detection, in particular to a pinhole camera shooting intelligent terminal detection technology based on ToF imaging.

Background

With the continuous upgrade of network communication technology and electronic optical technology, the miniature camera equipment is more and more concealed and more popularized. Some people use miniature devices for illegal activities such as stealing business secrets, stealing personal privacy, etc., and even forming a black industry chain. For example, 6800 many illegal camera-stealing privacy cases have been reported in korea alone within a year. The pinhole camera is small in size, and once the pinhole camera is hidden in places such as hotels, residential accommodations, market fitting rooms and the like, a professional detection tool is not available, so that the pinhole camera is difficult to find, personal privacy is greatly threatened, and therefore effective detection technical equipment is needed to check the places where shooting is not desired. Detection equipment on the existing market is rare in types, and detection devices based on infrared light reflection are common. The basic principle is that an LED is controlled to emit infrared light, and a light sensor receives ambient reflected light to form images. Because the reflectivity of the camera lens is significantly higher than the ambient environment, a pinhole camera can be positioned according to the reflected light intensity.

The existing detection device for hiding the camera can be divided into three types according to the working principle. The first category is radiation-based detection devices. The principle is that whether a hidden camera exists in a space is judged by capturing radiation characteristic analysis sent by the camera during working. The method has the disadvantages that the radiation of the camera is very weak, the camera is very easy to submerge in the environment, the acquisition and analysis difficulty is very high, and the detection reliability is poor. In addition, detection cannot be performed when the camera does not operate. The second type is a detection device based on WiFi signals, a part of hidden cameras need to transmit videos in real time through WiFi in the environment, and whether the hidden cameras exist can be detected by analyzing WiFi signal characteristics. This type of method can only detect wireless cameras, but cannot detect cameras with storage or transmitted over ethernet. The third category is optical-based detection devices, which are also currently more commonly used. The optical detection device detects the pinhole camera based on the reflection principle of visible light or infrared light, and compared with the first two devices, the detection principle of the devices is more reliable, and the application limitation is smaller. However, factors such as the spatial relative position of the detection device and the pinhole camera, and the mirror material objects distributed around the pinhole camera are likely to induce false detection; the existing LED laser detection device is easy to miss detection due to low imaging resolution; visible light detection is also susceptible to ambient light interference.

Disclosure of Invention

Technical problem to be solved

Aiming at the defects of the prior art, the invention provides a pinhole camera shooting intelligent terminal detection technology based on ToF imaging, and solves the problem that most of the existing methods for detecting pinhole cameras are not accurate enough.

(II) technical scheme

In order to achieve the purpose, the invention is realized by the following technical scheme: the utility model provides a pinhole intelligent terminal detecting system that makes a video recording based on ToF formation of image, includes:

intelligent detection terminal, it includes:

ToF shooting module and ToF image processing module

And the ToF shooting module collects images and sends the images to the ToF image processing module, and then a target detection algorithm is operated for detection.

Preferably, the image processing can be performed by a raspberry pi or a mobile computing terminal such as a smart phone.

Preferably, the target detection algorithm based on deep learning needs to train a network model, including using YOLOv5 target detection network:

acquiring ToF images of the pinhole cameras in different scenes, and manually marking the positions of the cameras to form a ToF image training set;

the process of training YOLOv5 on the training set until the maximum verification accuracy is reached can be completed on a computer, and then the trained model is transplanted to a ToF image processing module.

Preferably, the intelligent terminal detection method specifically includes the following steps:

s1, shooting a suspicious area by a ToF shooting module;

s2, the ToF image processing module inputs the ToF picture into a YOLOv5 network for detection, outputs a suspected target position range, marks the suspected target position range by using a candidate frame, outputs a detection confidence coefficient, and judges the ToF picture to be positive if the ToF picture is more than 0.5;

s3, if the visual angle of single shooting cannot completely cover the suspicious region, keeping the distance between the detector and the suspicious region approximately unchanged, keeping the orientation of the camera approximately unchanged, translating the detector in multiple steps for a proper distance in the horizontal direction or the vertical direction, repeating the step S1 and the step S2 every translation step, and according to a Z-shaped or S-shaped route until all the suspicious regions are covered.

(III) advantageous effects

The invention provides a pinhole camera shooting intelligent terminal detection technology based on ToF imaging. The method has the following beneficial effects:

the invention provides a pinhole camera intelligent detection terminal technology based on ToF imaging, which introduces a ToF imaging technology and a deep learning target detection technology based on the existing infrared detection principle, and provides a pinhole camera intelligent detection terminal to realize a detection effect with higher precision, wherein the basic flow is as follows: the method comprises the steps of shooting a suspicious area by using the ToF, analyzing and judging whether a pinhole camera exists or not by using a target detection technology after an optical photo is obtained, positioning, and enabling a detection result to be more reliable, the false detection rate and the missed detection rate to be remarkably reduced, the transportability to be good and the method to be deployed on a mobile terminal such as a raspberry pie with a ToF camera or a smart phone.

Drawings

FIG. 1 is a schematic diagram of pinhole camera detection according to the present invention;

FIG. 2 is a flow chart of the intelligent detection terminal of the present invention;

fig. 3 is a test experimental diagram of the pinhole camera in four scenes according to the invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Example (b):

as shown in fig. 1-3, an embodiment of the present invention provides a pinhole camera shooting intelligent terminal detection system based on ToF imaging, including:

intelligent detection terminal, it includes:

ToF shooting module and ToF image processing module

And the ToF shooting module collects images, sends the images to the ToF image processing module and then operates a target detection algorithm for detection.

A pinhole camera shooting intelligent terminal detection system based on ToF imaging can use a raspberry pi or an intelligent mobile phone mobile computing terminal to perform image processing.

The target detection algorithm based on deep learning needs to train a network model, and comprises the following steps of adopting a Yolov5 target detection network:

acquiring TOF images of pinhole cameras in different scenes, and manually marking the positions of the cameras to form a ToF image training set;

YOLOv5 is trained on the training set until the maximum verification accuracy is reached, the process can be completed on a computer, and then the trained model is transplanted to a ToF image processing module.

The intelligent terminal detection method specifically comprises the following steps:

s1, shooting a suspicious region by a ToF shooting module;

s2, inputting a ToF picture into a YOLOv5 network for detection by a ToF image processing module, outputting a suspected target position range, marking the suspected target position range by a candidate frame, outputting a detection confidence coefficient (between 0 and 1), and judging the ToF picture to be positive if the ToF picture is more than 0.5;

s3, if the visual angle of single shooting cannot completely cover the suspicious region, keeping the distance between the detector and the suspicious region approximately unchanged, keeping the orientation of the camera approximately unchanged, translating the detector in multiple steps for a proper distance in the horizontal direction or the vertical direction, repeating the step S1 and the step S2 every translation step, and according to a Z-shaped or S-shaped route until all the suspicious regions are covered. Fig. 3 shows detection effect diagrams of four hidden pinhole camera scenes, including a set top box, a bare pinhole camera, a wall clock, and a paper extraction box, where the left diagram is an actual position of the camera, the right diagram is a predicted position and confidence level output by an intelligent terminal, and an imaging light spot of the pinhole camera is accurately positioned and successfully distinguished from other light spots.

The ToF camera system obtains the depth information of the target by continuously sending light pulses to the target, has the advantages of quick response and low cost, and is widely applied to the 3D imaging function of the smart phone in recent years.

The ToF camera system comprises image sensor, image processing chip and modulation light source, and its theory of operation: the method comprises the steps of illuminating a scene by using a modulation light source, measuring the phase difference of reflected light waves, calculating the time difference of light flight through the phase difference or directly calculating the time difference of emitted light and reflected light, and calculating the distance between a camera and each point in the scene.

Unlike lidar, toF camera systems do not use laser beams to scan a space line by line, but rather one pulse illuminates the entire scene for imaging, thus allowing extremely high frame rates to be achieved, allowing 3D depth information to be extracted from the scene in real time using an embedded processor.

In addition, the ToF camera system has simple structure and low cost of main accessories, comprises an image sensor, a common laser generator and the like, and is very suitable for large-scale production and use.

The mirror surface material and the special shape of the pinhole camera cause the reflectivity of the pinhole camera to be very high, and when a ToF camera system shoots a suspicious region with the pinhole camera, under general conditions, the brightness of pixels corresponding to the lens position of the pinhole camera is obviously higher than that of other surrounding pixels, and the pinhole camera can be easily detected based on the characteristic.

Fig. 1 shows the principle of infrared light detection miniature commercial cameras, the hidden camera is generally composed of an electronic camera module and a hidden module, i.e. a shell, the camera module comprises two main components, namely an image sensor and a lens, the image sensor is mainly a Charge Coupled Device (CCD) or a Complementary Metal Oxide Silicon (CMOS) sensor, the reflectivity to light is high, the diameter of a pinhole lens is about 1 to 2mm, and most of the lens is designed into a round shape which is easier to process.

In a limited field of View (FoV), almost all incident laser beams are refracted by a lens and reflected by an image sensor, and can return to a light source along a direction opposite to the incident direction, which is also called "cat eye reflection".

The ToF camera system can then use these properties to detect the pinhole camera. It should be noted that successful detection requires the ToF camera system to be placed in the FoV and kept at a proper distance from the detected target, on one hand, the intensity of the reflected light is inversely proportional to the square of the distance, and on the other hand, the pinhole camera is small in physical size, and the resolution of the ToF imaging system is required to be higher the farther the distance is, so the effective detection distance depends on the power and resolution of the ToF.

FIG. 3 is a test experimental diagram of a pinhole camera in four scenes, wherein the left diagram is the actual position of the pinhole camera and marked as "pos it; and the right graph is the predicted position and confidence coefficient output after the intelligent terminal detects.

The target detection aims at positioning and identifying objects existing in images, in recent years, deep learning theories and technologies are developed, the supporting target detection is greatly developed, a target detection method based on a deep network needs to train a network model on a marked data set firstly and then perform detection judgment, the deep network has the advantages of learning and understanding high-level optical characteristics of a pinhole camera and distinguishing detailed characteristics such as shape, strength range and the like which are difficult to distinguish by human eyes, so that more accurate interpretation can be completed, the false detection rate and the false detection rate are effectively reduced, and the common target detection method based on the deep learning is generally divided into two types: the two-stage target detection method and the single-stage method firstly extract k candidate detection windows with unspecified categories, then further classify and regress the candidate detection windows to generate a final detection result, different from the two-stage method, the single-stage method directly classifies and regresses anchor points, generally speaking, the two-stage method has high detection precision, and the single-stage method has high reasoning speed, and particularly to an application scene of the invention, a ToF image reflects the intensity distribution of reflected light of a shot area, the scene content is simple, the single-stage method can achieve high detection precision, the requirement on computing power is low, and packaging and popularization are facilitated, so that the representative YOLOv5 network of the single-stage method is selected to analyze the ToF image, and other target detection networks can also be selected according to model situations.

Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (4)

1. The utility model provides a pinhole intelligent terminal detecting system that makes a video recording based on ToF formation of image which characterized in that includes:

intelligent detection terminal, it includes:

ToF shooting module and ToF image processing module

And the ToF shooting module collects images and sends the images to the ToF image processing module, and then a target detection algorithm is operated for detection.

2. The system for detecting the pinhole camera shooting intelligent terminal based on the ToF imaging as claimed in claim 1, characterized in that: the image processing can be carried out by using a raspberry pi or a mobile computing terminal such as a smart phone.

3. The deep learning based target detection algorithm according to any one of claims 1-2, which requires training a network model, and is characterized by comprising adopting a YOLOv5 target detection network:

acquiring ToF images of the pinhole cameras in different scenes, and manually marking the positions of the cameras to form a ToF image training set;

the process of training YOLOv5 on the training set until the maximum verification accuracy is reached can be completed on a computer, and then the trained model is transplanted to a ToF image processing module.

4. The intelligent terminal detection method according to claim 3, specifically comprising the steps of:

s1, shooting a suspicious region by a ToF shooting module;

s2, inputting a ToF picture into a YOLOv5 network for detection by a ToF image processing module, outputting a suspected target position range, marking the suspected target position range by a candidate frame, and outputting a detection confidence coefficient, wherein the ToF picture is judged to be positive if the ToF picture is larger than 0.5;

s3, if the visual angle of single shooting cannot completely cover the suspicious region, keeping the distance between the detector and the suspicious region approximately unchanged, keeping the orientation of the camera approximately unchanged, translating the detector in multiple steps for a proper distance in the horizontal direction or the vertical direction, repeating the step S1 and the step S2 every translation step, and according to a Z-shaped or S-shaped route until all the suspicious regions are covered.

Images