农业工程学报
農業工程學報
농업공정학보
2013年
23期
147-152
,共6页
彭江南%谢宗铭%杨丽明%孙宝启%王建华%孙群
彭江南%謝宗銘%楊麗明%孫寶啟%王建華%孫群
팽강남%사종명%양려명%손보계%왕건화%손군
种子%分选%棉花%种子物理指标%种子活力%Seed Identification软件
種子%分選%棉花%種子物理指標%種子活力%Seed Identification軟件
충자%분선%면화%충자물리지표%충자활력%Seed Identification연건
seed%separation%cotton%seed physical traits%seed vigor%Seed Identification software
为了研究Seed Identification软件在棉花种子加工工艺和精选参数选择上应用的可行性,以鲁棉28酸脱绒棉籽为材料,通过扫描仪获取400粒棉籽的PNG图像,利用Seed Identification软件快速提取图像中棉籽的RGB、Lab、HSB、灰度、长度、宽度和投影面积等物理指标,通过卷纸发芽获得每颗幼苗鲜质量作为种子的活力指标,种子物理指标与种子活力的相关性分析表明:幼苗鲜质量与R、S、B(HSB)、b、宽度、长度、投影面积的相关系数均达到0.05显著水平。按R<90、S≤18、B(HSB)≤36、b≤4、宽度>4 mm、长度>7.2 mm、投影面积≥25 mm2对种子进行精选,发芽率可由原来的89%分别提高到96.1%、95.1%、95.1%、95.3%、93.1%、93.5%、94.4%,获选率分别为96.6%、99.2%、98.9%、97.8%、98.6%、97%、94.7%。验证试验将种子按以上指标精选后,发芽率分别为95.1%、95.1%、94.8%、94.8%、94.4%、94.4%、94.8%。该研究为基于机器视觉技术对脱绒棉种实施快速、有效精选提供了理论依据。
為瞭研究Seed Identification軟件在棉花種子加工工藝和精選參數選擇上應用的可行性,以魯棉28痠脫絨棉籽為材料,通過掃描儀穫取400粒棉籽的PNG圖像,利用Seed Identification軟件快速提取圖像中棉籽的RGB、Lab、HSB、灰度、長度、寬度和投影麵積等物理指標,通過捲紙髮芽穫得每顆幼苗鮮質量作為種子的活力指標,種子物理指標與種子活力的相關性分析錶明:幼苗鮮質量與R、S、B(HSB)、b、寬度、長度、投影麵積的相關繫數均達到0.05顯著水平。按R<90、S≤18、B(HSB)≤36、b≤4、寬度>4 mm、長度>7.2 mm、投影麵積≥25 mm2對種子進行精選,髮芽率可由原來的89%分彆提高到96.1%、95.1%、95.1%、95.3%、93.1%、93.5%、94.4%,穫選率分彆為96.6%、99.2%、98.9%、97.8%、98.6%、97%、94.7%。驗證試驗將種子按以上指標精選後,髮芽率分彆為95.1%、95.1%、94.8%、94.8%、94.4%、94.4%、94.8%。該研究為基于機器視覺技術對脫絨棉種實施快速、有效精選提供瞭理論依據。
위료연구Seed Identification연건재면화충자가공공예화정선삼수선택상응용적가행성,이로면28산탈융면자위재료,통과소묘의획취400립면자적PNG도상,이용Seed Identification연건쾌속제취도상중면자적RGB、Lab、HSB、회도、장도、관도화투영면적등물리지표,통과권지발아획득매과유묘선질량작위충자적활력지표,충자물리지표여충자활력적상관성분석표명:유묘선질량여R、S、B(HSB)、b、관도、장도、투영면적적상관계수균체도0.05현저수평。안R<90、S≤18、B(HSB)≤36、b≤4、관도>4 mm、장도>7.2 mm、투영면적≥25 mm2대충자진행정선,발아솔가유원래적89%분별제고도96.1%、95.1%、95.1%、95.3%、93.1%、93.5%、94.4%,획선솔분별위96.6%、99.2%、98.9%、97.8%、98.6%、97%、94.7%。험증시험장충자안이상지표정선후,발아솔분별위95.1%、95.1%、94.8%、94.8%、94.4%、94.4%、94.8%。해연구위기우궤기시각기술대탈융면충실시쾌속、유효정선제공료이론의거。
In this paper, the correlation between the physical traits and seed vigor of delinted cotton seed (Lu Mian 28) was analyzed. The PNG format images of 400 cotton seeds were acquired with falatbed scanner, and the color features of cotton seed such as RGB, HSB, Lab, gray scale, and width, length and projected area were extracted automatically and quickly using seed identification software developed by our lab. Our seed identification software can identify the image and record related seed physical information, and then output all the information into an Excellfile automatically. The identifying results were achieved in 1 second, very quickly, with errors lower than 2%. The germination experiment was performed to get seedling fresh weight as seed vigor. Data analysis showed that R, H, S, B(HSB), b, width, length, and projected area had a relatively high coefficient of variation (more than 0.1) during the sample. Correlation analysis showed that R, S, B (HSB), b, width, length, and projected area were all significantly correlated with a seedling’s fresh weight. The correlation coefficients (R) were -0.128, -0.143, -0.121, -0.151, 0.283, 0.173, and 0.346 respectively. Cotton seeds of R﹤90, S≤18, B(HSB)≤36, b≤4, width>4mm, length>7.2 mm, seed projected area≥25 mm2 were selected respectively, and the seed germination rate was improved from 89% to 96.1%, 95.1%, 95.1%, 95.3%, 93.1%, 93.5% and 94.4%, and the selected rates of high quality seeds were 96.6%, 99.2%, 98.9%, 97.8%, 98.6%, 97%, and 94.7%, respectively. The verification test selected cotton seeds based on the physical traits and selected parameters described above, and the germination rate of seeds with R﹤90, S≤18, B(HSB)≤36, b≤4, width>4 mm, length>7.2 mm, seed projected area≥25 mm2 reached 95.1%, 95.1%, 94.8%, 94.8%, 94.4%, 94.4% and 94.8%, respectively. Therefore, we could deduce that this seed identification software could be applied to the selection of seed processing technology and the parameter determination of single delinted cotton seed according to seed vigor. The result has great importance for improving the seed processing level of China. In addition, our experiment confirmed that the seed projected area had a higher correlation with seed vigor compared to seed width and length. The reason may be that the projected area combined the information of seed width and seed length. But until now, there have been no related seed processing machines which could select seeds according to the seed projected area. This kind of seed machine is suggested to be developed as soon as possible.