| Abstrakt: |
In order to find out how well augmented recurrent neural networks operate for detecting when fruit is ripe in comparison to ridge regression, which has lower accuracy. Approach and methodology: Mango samples were collected, stored at room temperature, and checked for ripening, patches, and colour on a regular basis. While taking pictures of the samples, four fluorescent lamps were used, each with a color-rendering index (Ra) of over 95%. the MATLAB 6.5 algorithms for colour analysis, brown spot measurement, image texture analysis, and full-image preprocessing Group 1 (N=20 samples) was given an Enhanced Recurrent Neural Network (ERNN), while group I (N=20 samples) was given ridge regression. The parameters that were determined to define efficient fruit ripening detection were applied to 40 samples, 20 samples each group. Here, Group I serves as the control group, and Group II as the experimental group. We estimated the sample size using the following parameters: G power=80%, confidence level=95%, and alpha=0.05. Findings: It seems feasible to online anticipate the stages of mango ripening using computer vision. In terms of statistical significance, the two groups are distinct from one another. Statgraphics were used to quickly produce classification results. According to the results of the significance test, ERNN and Ridge regression were significantly different from one another (p=0.002). As a result, the produced significant gap is 0.002 (p<0.05). Conclusion: Using the csv (common separated values) and the hunterlab colorimeter, we were able to collect L*, a*, and b* measurements from 160 mango samples from the initial trial with an accuracy of 95.70. [ABSTRACT FROM AUTHOR] |
