Scholarlink

The care of the parameters in production processes in the food industry is becoming more demanding every day and requires that these be determined in the shortest possible time. In addition, it is preferred that the analysis methods are non-destructive and can be adapted to continuous processes (for process automation). The routine physical parameters used in fresh foods are weight, volume, brix density, pH, among others. This work developed a methodology for determining the weight and volume of fruits (mandarins and lemons) by means of artificial vision using free software. The image acquisition system consisted of a dark cabin, illuminated with diffuse light, a cell phone camera, ImageJ/Fiji software and a spreadsheet. For the training of the system, 80% of samples (100 mandarins and 100 lemons) of different sizes were used and for validation, 20%. Calibration, selection of geometric parameters (area, perimeter and length) and analysis of the images were done in ImageJ/Fiji(Escala de grises-canal azul). The predictive models for mass and volume of mandarin showed correlations higher than 94% with respect to their evaluated parameters: area, perimeter and average length. The area was the best responder (r \( ^{2} = 0.9754\) for volume and r \( ^{2} = 0.9567\) for weight). On the other hand, the predictive models for mass and volume of lemon showed correlations higher than 78%, with the area being the best responder (r \( ^{2} = 0.8855\) for volume and r \( ^{2} = 0.8764\) for weight). The area as a morphological parameter that had the best performance by demonstrating the best correlation would be the most suitable for predicting weight and volume in mandarins and lemons. These findings not only improve the understanding of physical relationships in fruits, but can also be applied in automated sorting and classification systems, optimizing industrial processes. The results obtained conclude that the methodology used “artificial vision with ImageJ” presents high precision, accuracy and reliability to be used in these or other fruits, thus allowing them to be used in continuous and automated systems (use of macros).

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Weight and Volume Prediction of Fresh Food (Fruits), by Artificial Vision: Imagej/Fiji

  • 章节

  • 发表日期：

    2025年

查看全文
  • Fernando Huamán Sánchez,
  • Mariano Luján Corro

摘要

The care of the parameters in production processes in the food industry is becoming more demanding every day and requires that these be determined in the shortest possible time. In addition, it is preferred that the analysis methods are non-destructive and can be adapted to continuous processes (for process automation). The routine physical parameters used in fresh foods are weight, volume, brix density, pH, among others. This work developed a methodology for determining the weight and volume of fruits (mandarins and lemons) by means of artificial vision using free software. The image acquisition system consisted of a dark cabin, illuminated with diffuse light, a cell phone camera, ImageJ/Fiji software and a spreadsheet. For the training of the system, 80% of samples (100 mandarins and 100 lemons) of different sizes were used and for validation, 20%. Calibration, selection of geometric parameters (area, perimeter and length) and analysis of the images were done in ImageJ/Fiji(Escala de grises-canal azul). The predictive models for mass and volume of mandarin showed correlations higher than 94% with respect to their evaluated parameters: area, perimeter and average length. The area was the best responder (r \( ^{2} = 0.9754\) for volume and r \( ^{2} = 0.9567\) for weight). On the other hand, the predictive models for mass and volume of lemon showed correlations higher than 78%, with the area being the best responder (r \( ^{2} = 0.8855\) for volume and r \( ^{2} = 0.8764\) for weight). The area as a morphological parameter that had the best performance by demonstrating the best correlation would be the most suitable for predicting weight and volume in mandarins and lemons. These findings not only improve the understanding of physical relationships in fruits, but can also be applied in automated sorting and classification systems, optimizing industrial processes. The results obtained conclude that the methodology used “artificial vision with ImageJ” presents high precision, accuracy and reliability to be used in these or other fruits, thus allowing them to be used in continuous and automated systems (use of macros).