Huimin GAO, Yu TIAN, Zhiqiang LIU, Yanpo CAO

Institute of Cash Crops, Hebei Academy of Agriculture and Forestry Sciences, Shijiazhuang 050051, China

Abstract Asparagus officinalis L. is favored by its high health function, but its hybrid seeds are expensive. The amount of seed, seed plumpness and germination rate are related to the production costs of breeding enterprises and large growers. Therefore, it is necessary to investigate the seed number and thousand kernel weight of A. officinalis L. This study developed a quick and accurate method to measure the seed number and thousand kernel weight of A. officinalis L. using image processing technology. Seed sample of A. officinalis L. was scanned with 200 dpi resolution, and the seed number was then obtained using Image-ProPlus software. After weighing the seeds, thousand kernel weight was finally calculated. By recording ‘macro’, the batch processing of the samples can also be realized. This method is simple and accurate, and can greatly save the time of investigation.

Key words Asparagus officinalis L., Seed number, Thousand kernel weight, Image processing method

1 Introduction

AsparagusofficinalisL. belongs toAsparagus, Liliaceae Juss., and it is dioecious perennial plant. Its young stalk is edible, with comprehensive nutritional value.A.officinalisL. is rich in protein, amino acid, microelement and flavonoids, with the effects of reducing blood pressure and fat, detoxifying, diuresis, and enhancing immunity. The extract ofA.officinalisL. contains steroidal saponins, with the function of preventing cancer cell proliferation, and it has good effect on many kinds of cancers. So it is called "the king of vegetables" and is defined as the first of top ten vegetables in the world by the World Health Organization[1-6]. With deep understanding on value ofA.officinalisL., its market demand is increasing, with good development prospect. At present, China has become a major producer ofA.officinalisL. in the world, and its planting area is 0.1 million ha, occupying the first place in the world[7-8]. Most seeds ofA.officinalisL. are imported, and price of hybrid seeds is 8 000-10 000 yuan/kg, which is expensive[9]. The amount of seed, seed plumpness and germination rate are closely related to the production cost. Therefore, it needs to investigate seed number and thousand kernel weight ofA.officinalisL. Traditional seed counting and thousand kernel weight measurement both depend on manual work, which consumes time and labor, and has not been suitable for the development of modern agriculture. Based on image acquisition and digital processing, a quick, simple and accurate method measuring seed number and thousand kernel weight is developed, to improve production efficiency and save labor cost.

2 Materials and methods

2.1 MaterialsThe variety "Jilu Luwang" was selected, and its seeds were full, without obvious diseases and insect pests. It was provided and conserved by Institute of Cash Crops, Hebei Academy of Agriculture and Forestry Sciences.

2.2 Instruments and equipmentScanner (EPSON V370), and it was set as "color photo" mode, with 200-800 dpi of resolution. Conventional computer, and Image-Pro Plus V6.0 image processing software was installed (download address: http://www.mediacy.com).

2.3 Methods

2.3.1Determination of scanning resolution. Resolution of scanner was respectively set as 200, 400, 600 and 800 dpi, and the randomly selected seeds ofA.officinalisL. were scanned, to record scanning time and size of storage file. Finally, the optimal scanning resolution was determined.

2.3.2Manual test method. Artificial measurement method of thousand kernel weight ofA.officinalisL. seed referred to theInspectionofGrainandOilseedsMethodsforDeterminationoftheMassof1 000Grains(GB5519-85). Seeds ofA.officinalisL. were counted from sample without impurity, and weighing was conducted by the balance, and value was recorded (accurate to 0.1 g). Manual counting was conducted, which was repeated for 3 times. Thousand kernel weight was calculated, and its mean was taken.

2.3.3Determination of seed number by image processing method. Seeds ofA.officinalisL. were loosely spread on the plate of scanner for scanning (Fig.1a), and photo was conserved. Image-pro plus 6.0 software was used to open the conserved photo, and then "count/size" button of tool bar was clicked. After that, "select colors" button in the pop up dialog box was clicked. Straw was used to select seed color, and most of seeds were selected (Fig.1b). The dialog box was closed, and "count" button in "count/size" dialog box was clicked (Fig.1). After that, "statistics" button of view in menu bar was clicked, and number of seed could be directly obtained. The result could be verified and confirmed by artificial counting method.

Fig.1 Determination of seed number using image processing

2.3.4Thousand kernel weight ofA.officinalisL. seed. Seed sample was weighed by analytical balance (0.01 g), and sample weight and seed number were set in the formula (1) to calculate thousand kernel weight. It was repeated for three times, and mean was taken.

Thousand kernel weight of sample=Sample weight/Seed number×1 000

(1)

3 Results and analyses

3.1 Comparative analysis of scanning resolutionImage-Pro Plus software was used to process the images of the randomly selected seeds with 4 kinds of resolutions after scanning. Seen from Table 1, image resolution did not affect Image-Pro Plus software, and all of them can count the samples accurately. There was larger difference in scanning time and storage space of image file. With resolution was improved, scanning time increased from 12 to 98 s, while image file also increased from 0.29 to 3.12 Mb. Therefore, considering working efficiency, 200 dpi of resolution was used.

Table 1 The effect of scanning resolution on seed number counting

Scanningresolution∥dpiScanningtime∥sImagefilesize∥MbImageanalysiscounting∥grains200120.29256400230.78256600321.51256800983.12256

3.2 Comparative analysis between image processing method and traditional artificial methodImage processing method and artificial counting method were respectively used to count the sameA.officinalisL. sample, and time was recorded. Seen from Table 2, image processing method could accurately countA.officinalisL. seed, and the time required was obviously less than artificial counting method. It is because that image processing method was based on image scanning and software identification calculation, and seed number had little impact on time, while the time required by artificial counting method increased significantly with seed number. It illustrated that image processing method had high accuracy and could significantly improve working efficiency.

Table 2 Time-consuming comparison between image processing method and traditional manual method

A.officinalisL.sampleImageprocessingmethodCountinggrainsTimesManualcountingmethodCountinggrainsTimes11582015852236221362127358523585212

Fig.2 Image batch processing by recording a ‘macro’ file

3.3 Multiple-sample measurement by image batch processingIf sample amount is large, scanning-identification calculation one by one of single sample could affect working efficiency. In the experiment, batch and rapid processing of image could be realized by recording "macro". First, image scanning of all samples was conducted. After scanning of all samples was completed, the recorded "macro" was used for identification calculation of image file, which could greatly improve measurement velocity. Operation process of "macro" record was as below: opening a picture to be analyzed, clicking "Macro"-"Record Macro" button of menu bar, inputting a name to "macro" in the pop up "Record Macro" dialog box, such as "measurement of thousand kernel weight ofA.officinalisL., and setting a shortcut key for the macro in "Shortcut Key", such as "Ctrl+A" (Fig.2). After the process of image processing and analysis was recorded into a macro file, batch analysis of multiple images was conducted by the macro file, and seed number ofA.officinalisL. could be obtained by directly pressing shortcut key "Ctrl+A".

4 Conclusions and discussions

Seed quality has certain impact on yield and quality of vegetable. Before sowing, cleanness of seed, thousand kernel weight, germination percentage and germination potential should be determined. In sowing process, plantation density and seed use amount should be controlled accurately. In this paper, via scanner imaging and image processing with image analysis software, quick and accurate measurement of seed number and thousand kernel weight ofA.officinalisL. was realized, and working efficiency was improved, which decreased subjective deviation caused by artificial counting, and saved human and material resources. In recent years, image processing method was applied in crop character investigation and field data collection. For example, Xiao Yongguietal.[10]used image processing method to quickly and effectively carry out quantitative analysis on vegetation coverage in seedling and booting stage. Xiao Qiangetal.[11]used digital camera and Photoshop software to determine leaf area of plant, without destruction. Wei Quanquanetal.[12]carry out nitrogen nutrition diagnosis of winter rape based on digital image technology. Wang Fangyongetal.[13]used digital image to estimate proportion of photosynthetic effective radiation absorption of cotton. With the development of science and technology, information technology develops rapidly, and scanner, digital camera, smart phone, tablet computer and other digital image acquisition tools have been widely used in people’s lives. The performance, cost performance and operability of such image devices are constantly improving, and the price is also getting lower. It is no longer difficult to apply image processing technology to the investigation of crop characters, which will greatly improve the working efficiency and data accuracy of field characters investigation in crop breeding. In this paper, it was verified that it was feasible to count seed number ofA.officinalisL. and measure thousand kernel weight by image processing method. This method is simple, quick and accurate, and is ideal method of counting seed and measuring thousand kernel weight.