Overload intelligent detection method of lift platform’s number of people under auxiliary visual

Yue PANG

(Chongqing Vocational Institute of Engineering, Chongqing 402260, China)

?

Overload intelligent detection method of lift platform’s number of people under auxiliary visual

Yue PANG*

(ChongqingVocationalInstituteofEngineering,Chongqing402260,China)

In view of the overload detection of current lift platform’s number of people is mainly manual work, which has man-made interference and its disadvantages are very obvious, a lift platform based on computer vision image statistical method was put forward. Discrete wavelet transform algorithm was used to de-noise the collect images of lift platform and the Gaussian curve fitting method was used to have image feature extraction. The change of head and the grey value of contour area of background are continuous. There are differences between the texture features of workers contour and the contour of a background. And the gray value of workers contour edge pixel has gradient change. We divided the head and marked and counted the characteristics of personnel area according to segmentation result. The experimental results show that the method used can have statistics on the number of people on the elevator platform accurately and its error is less than 5%, which can meet the demand of practical application.

Elevator platform, People counting, Computer vision

People counting is the key of the monitoring and control system of lift platform, and the number of workers on lift platform is the very important factor to the personal safety of workers and the safe operation of it[1-2]. Lift platform has very strict requirements in the number of workers, which always be neglected. Therefore it is necessary to have an accurate accounting of workers on the platform to avoid the overload operation and to make sure the personal safety of workers and the safe operation of lifter. Manufacturers can set the “overload protection” according to the load capability and self security feature of the lifter: when the weight of workers is over the rated weight, the audio will have an alarm sound and the lifter cannot run in order to ensure the personal safety of workers[3-4]. Now the detection of overload of lifter is mainly manual work and will be influenced by human factor, which has obvious disadvantages and lowers the accuracy of people counting on the platform [5].

1 Preprocessing method for image acquisition of lift platform features

1.1Remoteimagetransmissionprincipleofliftplatform

The image acquisition equipment of the lift will have the collected image compression first, and then compression image will be remotely transmitted to the security monitoring system [6-7]. The steps are as follows.

Suppose that the number of workers on the platform lift isN, and workers image transmission distance isl, and the initial moment of image transmission ist0, and reaches moments of image to the security monitoring system ist1, and if the number of pixels of the arbitrary one frame of imageRiisM, and the gray value of it isSi, the image data compression process can be calculated through formula (1):

(1)

The above formula can calculate the image compression coefficient and can describe the degree of image compression.

Image resolution in security monitoring system can use formula (2) to calculate:

(2)

The above formula can calculate the resolution of the image, and can have the description of image definition.

1.2 Disadvantages of long-distance image transmission

From formula (1), we can see that the image compression coefficient decreases with the increase of the distance. Through formula (2), we can conclude that the decrease of image compression coefficient will reduce the resolution of the image in the security monitoring system. Because of the long distance between the elevator and security monitoring system, image will be affected by other factories interference source in the transmission process and the clarity of image will be reduced by compression, and therefore it contains a lot of noise. The traditional denoising methods are mainly image transmission method based on neural network algorithm and image transmission method based on ant colony algorithm and so on, whose denoising effect is not ideal, which leads to the poor clarity of image [8-9]. A new denoising method of image based on discrete wavelet transform was proposed.

Fig.1 Denoising effect of traditional algorithm

1.3Denoisingofliftplatformimage

Due to the long distance of communication and great external interference factors in the transmission process, the transmission of images of workers on the lift platform exists a large number of random noise, which leads to that staff in security monitoring system cannot have exact statistics of the number of workers in lift. After the Image being transmitted to the security monitoring system, the expansion of it is an image sequence, which is represented byg(y). Using the formula (3), we can calculate the transformation of the image sequence:

(3)

If the parameters in the formula satisfyk≥k0, the following formula can be obtained:

(4)

g(y),γk0,l(y) andφk,l(y) in the above formula are all functions of the discrete variables. Collect arbitrarily sampley0,Δy,y=1,2,…,N,g(y)=g(y0,yΔy)，and setk0=0，N=2′, and theny=1,2,…,N,k=1,2,…,Kandl=1,2,…,2′-1 can be summed up. Arbitrarily collect an image, and select four pixelsg(0)=1,g(1)=4,g(2)=-3 andg(3)=0, and sum upy=1,2,3,4，k=1,2, and substitute the above four pixels into the above formula, we can get the following formula:

(5)

Using the above formula, we can sample the image features of the workers on the sample. According to the formula below, the sample can be collected from the similar interval:

(6)

According to the formula below, we can have discrete wavelet transform:

(7)

The method of discrete wavelet transform can remove the noise from outside factor and eliminate the noise caused by external factors, thus to improve the contour articulation of the image.

Fig.2 Comparison of the denoising effect of the proposed algorithm and the traditional denoising method

2 Personnel statistics based on image features

2.1Gausscurvefittingprocessofpersonnelhead

In the process of image processing, the computer is mainly based on the contour of workers. Contour feature is the discontinuity of gray value between the background and the target image. The main features of the contour of the lift platform are: ① the gray value of the background contour is continuous; ② the texture features of the contour of workers and the texture of the background are different; ③ the gray value of the edge pixels of the workers’ contour has gradient changes.

Removing the noise of image acquisition can improve image clarity and improve the contour feature of the workers then. Using Gaussian curve fitting method, we had workers image processing, a Gaussian curve by deformation is as shown in equation (8):

(8)

Set the curve of second degree isz=By2+Cy+D, and then the gray value of pixels in the image of the workers can be calculated by using the formula (9):

(9)

In Fig.3, we set the maxima of gray scale difference value is 0 point, with the corresponding gray value isg0, and set the approaching two pixels between -1 and 1, and the two corresponding to the pixel gray value areg-1andg1, then the following formula can be used to describe the three pixel gray value:

(10)

g0=1/12B+D

(11)

g1=26/24B+C+D

(12)

The values of pointB,C,Dcan be calculated with the above formulas:

D=g0-1/12B=12/13g0-1/24g-1-1/24g1

With the formula (13), the space position parameters of top point can be calculated:

(13)

After the logarithm transformation of the formula, as shown in the formula (14), the spatial position parameters of worker images can be obtained:

(14)

By calculating each of the pixels of the workers’ image area, we can get a clear outline of the workers’ images, as shown in Fig.4.

Fig.3 Curve fitting outline of worker images

Fig.4 Gauss curve fitting processing image

2.2Peoplecountingandpositionmarking

After the processing of statistics and labeling of the outline, the statistical ideas used in this paper are as follows:

1) The idea of the people counting of the image of a frame lift platform is: inputting the image, marking the number and outputting the image. We set these three array which are, Inimage represents inputting the image, Markimage represents a marker image (where 0 is the background, 1 is the aim), Outimage represents outputting the image (where 1 is the first goal, and 2 is the second goal, and by parity of reasoning) respectively.

2) A scan was carried out from the periphery of an image, and tracked when scanned target of 1. In the scanning process, taking the boundary points of the contour as the target point to scan and if it found area points on the surrounding, it kept them in Outimage, while if it did not find pixel whose value is 1 around, the mark is over. After the end of the tag, it will continue to scan the remaining region for the pixels whose median is 1. When it scans to the new area, the value of Outimage will be increased by 1, so the target quantities can be expressed by Outimage.

3) Using the method (2) to scan the current image and end the tag. The number and position of the head of each person were marked. The results of the tag are shown in Fig.5.

Fig.5 Tag results

3 Experimental simulation results analysis

In the transmission process, the clarity of the image might be reduced and the image could contain a lot of random noise after compression. The effect of traditional denosing method is poor. Therefore, we suggested a method to clear statistic of people number on the lift platform based on discrete wavelet transform algorithm, in which the discrete wavelet transform algorithm was used to deal with the noise and eliminate the noise to provide accurate number foundation for the extraction of workers contour features. The image pixels of the workers were calculated by computer vision image technology and the clear portrait was obtained, which can realize speed statistics of the number of workers on the lift platform.

In order to verify the superiority of the algorithm proposed in this paper, a simulation experiment was needed. Experiments were carried out by simulating the lift platform. Equipment used in the tests include: the types of ag-hp x 250 of the camera produced by Matsushita, digital image acquisition card MV-1394B produced by the Microvision production models. And test configuration of the computer in this test includes: Intel Pentium Dual Core E5300 2.6 GHz CPU, 2G DDR3 memory. The image size is 320 * 240. When the test is fixed, the camera is placed in the appropriate height, and the area of the region is similar to the lift platform, and the image is processed by the time cut of every five frame. In the experiment, the images of the transmission of the traditional algorithm were compared. The results are shown in Fig.6.

Fig.6 Experimental contrast effect chart

Through the comparison of the images that the algorithm presented in this paper for image denoising processing can avoid the disadvantages of traditional image and can realize fast and accurate statistics of the number of workers on the platform lift. To analyze the experimental data, we summarized the following table:

Table 1 Comparison between the algorithm in this paper and the traditional algorithm

NumberofPeopleinLiftProposedAlgorithmDetectednumberofpeopleAccuracyrateTime-consuming/msTraditionalalgorithmDetectednumberofpeopleAccuracyrateTime-consuming/ms22100%692100%7644100%814100%11266100%132583%18688100%197787.50%26910990%243880%377121191.60%3051083.40%454424095.30% 3685.70%

From the comparison of diagram and experimental comparison chart, we can see that this algorithm is very good to solve the noise problem and can clear and accurately have the number of workers statistical on lift platform, which avoids the disadvantage of traditional algorithm, that the resolution is not high, poor accuracy and drawbacks. In the case of the time consuming and the traditional algorithm, the accuracy of the statistic is greatly improved (95.3%). The results show that the algorithm is of higher accuracy and has strong practicability.

4 Conclusions

This paper presented a statistical method for the people number counting of lift platforms based on computer visual images. Using discrete wavelet transform to denoise the image collection and improve the image definition, using Gaussian curve to calculate contour pixels one by one, we got a clear outline, and finally marked outline and count, and can have timely and accurate statistics on the number of workers on the platform lift to timely judge if the number of workers on lift platform is overloaded in order to protect the personal safety of workers, which is of high practical value.

[1]CHEN T H, CHEN T Y, CHEN Z X．An Intelligent People -Flow Counting Method for Passing through a Gate[C]//Proceedings of IEEE Conference on Robotics, Automation and Mechatronics Bangkok, Thailand,2006：1-6．

[2]IRSHAD A, MATTHEW N．Multiple Human Tracking in High Density Crowds [J]. Image and Vision Computing,2012,30:966-977．

[3]YING Jun, LI Qiang. Counting algorithm for numbers of people in elevator car based on improved Hough transform. Computer Engineering and Applications, 2014, 50(19)：165-168.

[4]YING Jun, LI qiang. Based on the improved Hough transform inside the elevator car number statistics algorithm [J]. Computer engineering and application, 2013,10.3778 / j.i SSN. 1002-1002. 1211-0053.

[5]JIN Haiyan，XIONG Qingyu, et al．Number Statistical Method within Elevator Car Based on Image Processing [J]. Chinese Journal of Scientific Instruments, 2011 ，32(6)：161-165．

[6]Jin HaiYan, Xiong Qingyu et al. Based on image processing of the elevator car number in statistical method research [J]. Journal of instruments and meters, 2011, 32 (6) : 161-165.

[7]LI Guangjie. A Face Data Nuclear Characteristics of the Complex Network Anomaly Detection Algorithm [J]. Bulletin of Science and Technology, 2013,8: 27-29.

[8]Li Guangjie. A data-oriented nuclear characteristics of complex network anomaly detection algorithm [J]. Science and technology, 2013, 8: 27-29.

[9]BORGEFORS G. Distance Transformation in Arbitrary Dimensions [J].Computer Vision，Graphics and Image Processing, 1984, 27：321- 345.

[10]VINCENT L, SOILLE P. Watershed in digital space: An Efficient Algorithm Dased on Immersion Simulation [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence，1991,13: 583-598.

[11]WU Zuoling, CHEN Xiong, et al. Palm-Dorsal Vein Biometrics Based on Watershed Algorithm [J].Microelectronics & Computer, 2008, 27(10)：114-119.

[12]Wu Zuoling, Chen Xiong et al. A kind of near infrared hand vein image recognition algorithm [J]. Journal of microelectronics and computers, 2008, 27 (10): 114-119.

[13]LI Y Y, SHI H Z, et al. Quantum Evolutionary Clustering Algorithm based on Watershed Applied to SAR Image Segmentation[J]. NeuroComputing, 2012, 87: 90-98.

(Continued from 105 page)

摘要：采用Pro/e軟件進行隔膜泵動力端-曲柄滑塊機構(gòu)的三維建模，利用機械系統(tǒng)動力學(xué)仿真軟件ADAMS構(gòu)建虛擬樣機，通過樣機仿真確定隔膜泵曲柄滑塊機構(gòu)的運動學(xué)和動力學(xué)參數(shù)，為隔膜泵動力端的設(shè)計打下基礎(chǔ)。

關(guān)鍵詞： Pro/E建模； 隔膜泵； 虛擬樣機；動力仿真

10.3969/j.issn.1001-3881.2015.24.022 Document code: A

TP391.41

輔助視覺下升降機平臺人數(shù)超載智能檢測方法

龐玥*

重慶工程職業(yè)技術(shù)學(xué)院, 重慶402260

針對當前的升降機平臺人數(shù)超載檢測以人工為主，存在人為因素干擾、弊端較為明顯的情況，提出一種基于計算機視覺圖像的升降機平臺人數(shù)統(tǒng)計方法。利用離散小波變換算法對采集的升降機平臺圖像進行去噪處理，通過高斯曲線擬合的方法進行人員頭部圖像特征提取，再根據(jù)人員頭部與背景輪廓區(qū)域灰度值的變化是連續(xù)性的、工人輪廓與背景輪廓的紋理特征有差異性、工人輪廓邊緣像素的灰度值存在梯度變化這3大特征，對人員頭部進行分割，根據(jù)分割結(jié)果對人員特征區(qū)域進行標記和計數(shù)。結(jié)果表明：利用該方法能夠準確地對升降機平臺上的人數(shù)進行統(tǒng)計，誤差低于5%，符合實際應(yīng)用需求。

升降機平臺；人數(shù)統(tǒng)計；計算機視覺

隔膜泵動力端虛擬樣機的計算機仿真研究與驗證

程志銘*,閆波,趙小飛

山西機電職業(yè)技術(shù)學(xué)院， 山西 長治046011

20 March 2015; revised 13 June 2015;

Yue PANG,

E-mail: 93439232@qq.com

accepted 15 October 2015

Hydromechatronics Engineering

http://jdy.qks.cqut.edu.cn

E-mail: jdygcyw@126.com