I. Introduction

In recent years, the development nuclear radiation detection based on CMOS technology has notably increased. However, improvements are still necessary to deploy accurate remote measurement systems for low-energy radiation. In fact, the traditional CMOS radiation detection technology is limited for remote detection, as the electro physical characteristics of the detector single crystals, CdZnTe and CdTe, and radiation monitoring of the gaseous electron multiplier muon detectors difficult remote measurements and demand high amounts of power ^[1-3]. In contrast, ZigBee modules are affordable, have low power consumption, and simplify networking and remote communication, and hence they can be suitable for remote measurement systems of low-energy radiation ^[4-5]. Unlike the traditional radiation detection systems, we propose the combination of an ARM board and ZigBee wireless transmission to achieve remote measurement of low-energy radiation. Moreover, we evaluated different methods to improve the calculation of the gray-scale value and developed the subtraction–summation algorithm based on gradient multiplication and edge detection. Consequently, we improved the measurement accuracy of low-energy radiation. Furthermore, the combination of ARM technology and ZigBee communication not only allows the remote measurement of low-energy radiation, but also improves the reliability and stability of the system.

II. Experimental setup

The block diagram of the system hardware is illustrated in Figure 1. We used the PA2 and PA3 pins from the serial port of the STM32F103 board. In the transmitter, the data (highlight positions and sum of the corresponding pixel gray-scale values) are processed in MATLAB and stored in the STM32F103 board, and then sent through the STM32F103 serial port 2 (pins PA2 and PA3). Specifically, for data processing, the data is transmitted through the external RS232 interface of pins PA2 and PA3. Then, the STM32F103 board and ZigBee module communicate through an interface that converts the RS232 signal into TTL communication (Tx–Rx, Rx–Tx, GND, Vcc), and the received data is stored in the ZigBee module. Finally, an LCD in the ZigBee module displays the coordinates and measurement outcomes.

Figure 1.Full-screen View

System hardware diagram.

III. Materials and methods

We first analyzed the max gray-scale value algorithm and found that it does not contain all the pixel gray-scale values of the highlights. Next, we analyzed the subtraction–summation algorithm that consists of three steps: 1) Image graying; 2) Background removal from the gray-scale image; 3) Gray-scale value sum of the whole image. We found that in the subtraction process, the error to determine the background gray-scale value affects the calculation accuracy and does not produce suitable outcomes.

Hence, we evaluated the whole summation algorithm that has a higher calculation accuracy compared to both the max gray-scale value and the subtraction–summation algorithms. The implementation of this algorithm is detailed as follows.

1) Image graying. According to the YUV color space, the Y component corresponds to the brightness of a pixel. The relationship between the RGB and YUV color space allows to establish a transformation from the RGB color components to the Y brightness component as

This brightness value is used as the gray value of the pixel, which is denoted as x_i.

2) Binarization of the gray image. We considered 256 brightness levels in the gray image, and the binarization threshold was set using the MATLAB automatic threshold selection function. The resulting binary image still reflects general and local features ^[8]. In this step, the gray-scale value corresponding to each pixel is denoted as y_i.

3) A new image is obtained by multiplying the binary and gray-scale images, and each resulting pixel value is denoted as z_i.

4) Finally, each pixel value of the multiplied image is summed up. The sum of the pixel gray-scale values for overall highlights can be determined by

After applying this algorithm, we found that it is sensitive to noise, and hence more suitable for applications having either weak or no noise.

Furthermore, the three above mentioned algorithms have a limited anti-interference ability, and the algorithm taking an n×n square as the boundary ^[9] presents some accuracy errors. Thus, we used edge detection and summed the pixel gray-scale values within the boundaries to introduce a method for calculating the radiation dose based on the gray-scale value of the corresponding images. Traditional edge detection algorithms are based on the first derivative to improve robustness against noise, but the edge location is not accurate. In contrast, other edge detection algorithms based on the second derivative result very accurate, but they are sensitive to noise. Hence, we propose a method based on gradient multiplication for edge detection. This method suitably and accurately locates the edges and is robust against noise, and the resulting detection provides more complete edges than traditional algorithms. We first employed the whole summation algorithm based on gradient multiplication and edge detection to improve the calculation method of the binarization threshold. For this method, the background gray-scale value is determined using the following process: 1) Determine the image gradient map using the Sobel operator; 2) Determine the image gradient map using the Laplacian operator; 3) Multiply the two maps to obtain another gradient map; 4) The image obtained from the new gradient map is binarized and then used as mask image; 5) Determine the background gray-scale value by obtaining the average gray-scale value of the edge pixels. The mask image in the fourth step and the original image were multiplied to obtain an image that only retains the gray-scale values of edge pixels. The sum of the gray-scale values in this image and the mask image are denoted by P and Q, respectively. Thus, the background gray-scale value is given by B=P/Q, which is considered as the binarization threshold. This threshold mitigates the error related to that from the original whole summation algorithm, and improves the anti-interference ability and consequently the stability of the system.

However, this algorithm does not remove the background gray-scale value, and thus generates error in determining the radiation dose. To solve this problem, we propose using the subtraction–summation algorithm based on gradient multiplication and edge detection. This algorithm improves the method for determining the background gray-scale value in the second step of the above mentioned subtraction–summation algorithm. In fact, for this improved algorithm, we use the background gray-scale value obtained from the whole summation algorithm based on gradient multiplication and edge detection. Consequently, we obtained higher accuracy, better consistency to determine the radiation dose, stronger anti-interference ability, and easier operation if compared to all the other algorithms mentioned in this section. Hence, we used the subtraction–summation algorithm based on gradient multiplication and edge detection to implement the proposed system.

IV. Results and discussion

A. Analysis of radiation image algorithms

Compared with both the max gray-scale value and the subtraction–summation algorithms, the whole summation algorithm features characteristics such as high speed, accurate positioning, low error, and simple operation. First, the gray-scale image is obtained by applying image graying. Next, the binary image is obtained from the gray image. Finally, another image is obtained by multiplying the binary and gray-scale images. In the original image, the bright spot cannot be clearly observed, whereas it is clearly visible after the image processing. A three-dimensional view of the image pixels can help to better visualize the location of the bright spot. The gray-scale image before processing is shown in Figure 2a, whereas the corresponding result after applying the whole summation algorithm is shown in Figure 2b. When the CMOS image sensor acquires an image, a γ photon usually hits several neighboring pixels and forms an irregular shape, and hence causes an error in the max gray-scale value algorithm. However, the whole summation algorithm can consider the gray-scale value of the whole photon, thus avoiding leakage and improving the measurement accuracy.

Figure 2.Full-screen View

Three-dimensional pixel intensity image corresponding to the radiation image of ⁶⁰Co. a Original radiation image of ⁶⁰Co. b Image processed using the whole summation algorithm.

We also compared the image processing capabilities between the max gray-scale value, subtraction–summation, and whole summation algorithms. For this comparison, we applied the algorithms to the radiation image of the same cobalt source acquired over 9 consecutive frames. The normalized gray-scale values for the images are shown in Table 1. A fourth-degree polynomial curve fitting was used in the data of Table 1, whose results are shown in Figure 3. The curves clearly describe the variation of radiation dose over time. In addition, it is clear that the curve corresponding to the max gray-scale value algorithm is very wide, which indicates a high sensitivity and causes a poor anti-interference ability. In contrast, the curve corresponding to the subtraction–summation algorithm is very narrow, which indicates a low sensitivity and a difficulty to detect small changes in radiation dose. Finally, the curve corresponding to the whole summation algorithm lays between those of the two other algorithms. Thus, this algorithm can detect small changes and has a suitable anti-interference ability in the presence of either weak or no noise.

Table 1.Full-screen View

Normalized gray-scale value of a 60Co source over consecutive frames

Algorithm	Image 1	Image 2	Image 3	Image 4	Image 5	Image 6	Image 7	Image 8	Image 9
Max gray-scale value algorithm	0.095	0.181	0.107	0.116	0.162	0.110	0.070	0.044	0.115
Subtraction–summation algorithm	0.055	0.230	0.076	0.149	0.305	0.070	0.023	0.001	0.091
Whole summation algorithm	0.070	0.209	0.079	0.157	0.243	0.077	0.055	0.016	0.094

Show more

Figure 3.Full-screen View

Curve for the data in Table 1.

Given the limited anti-interference ability of the whole summation algorithm, we used the subtraction–summation algorithm based on gradient multiplication and edge detection. To verify this method, we processed the same image using the original and improved versions of both the subtraction–summation and whole summation algorithms. The corresponding background gray-scale value and the binarization threshold are listed in Table 2.

Table 2.Full-screen View

Background gray-scale value and binarization threshold of the original and improved versions of the subtraction–summation and whole summation algorithms

Parameter (algorithm)	Original algorithm	Improved algorithm
Binarization threshold(whole summation algorithm)	6.018	3.152
Background gray-scale value (subtraction–summation algorithm)	3	3.152

Show more

Next, we processed the radiation detection images of the same cobalt source acquired over 9 consecutive frames using the subtraction–summation and whole summation algorithms, both based on gradient multiplication and edge detection. The original and min–max normalization data of the cobalt source radiation are listed in Table 3, and we employed a fourth-degree polynomial curve fitting for the normalized data in Table 3, which is shown in Figure 4. From the graph, we can see that the amplitude obtained from the subtraction–summation algorithm is higher than that obtained from the whole summation algorithm. Moreover, this amplitude shows an increasing value over time, which is consistent with the change of radiation dose. In addition, the subtraction–summation algorithm shows higher precision and robustness against noise. Furthermore, it suitably meets the design requirements of the ZigBee-based hardware and can retrieve more stable and reliable measurement data for the system.

Table 3.Full-screen View

Original and normalized data of a ⁶⁰Co source radiation over consecutive frames

Algorithm	Image 1	Image 2	Image 3	Image 4	Image 5	Image 6	Image 7	Image 8	Image 9
Improved subtraction–summation algorithm	14,788	7533	14,798	11,488	12,452	14,137	13,680	13,975	16,506
Min–max normalization	0.809	0.000	0.810	0.441	0.548	0.736	0.685	0.718	1.000
Improved whole summation algorithm	25,822	13,819	25,805	23,488	27,244	23,041	35,529	41,338	30,186
Min–max normalization	0.436	0.000	0.436	0.351	0.488	0.335	0.789	1.000	0.595

Show more

Figure 4.Full-screen View

(Color online) Curve for the data in Table 3 (Data 1 and Data 2 represent the normalized data from the subtraction–summation and the whole summation algorithms, respectively. Both algorithms were based on gradient multiplication and edge detection).

V. Summary

We designed and implemented a remote radiation-detection system based on an ARM board and a ZigBee module. In addition, we verified that the subtraction–summation algorithm based on gradient multiplication and edge detection improves the measurement accuracy. Then, we combined the data processing capabilities of the ARM STM32F103 board and the remote transmission function of the ZigBee module to perform remote data transmission of undistorted measurements of gamma waves irradiated from a cobalt source. The experimental results verified the effectiveness of employing the proposed algorithm and the feasibility of the system that combines ARM and ZigBee architectures. Furthermore, the system can be used for accurate and stable remote measurements to monitor low-energy radiation.