Cherry tomato firmness detection and prediction using a vision-based tactile sensor

Since fresh cherry tomatoes have a high likelihood of spoilage during post-harvest transport and storage, assessing the fruit grade based on its firmness or predicting its potential storage time becomes necessary. The study proposes a method for non-destructive fruit firmness detection using a photometric stereo vision-based tactile sensor. The tactile sensor can measure the three-dimensional deformation of cherry tomatoes by mapping between pixel color and surface normal using a gradient prediction network. The summation of the deformation heights measured by a tactile sensor pressing on the cherry tomatoes can accurately describe fruit firmness, which was also verified by a comparative experiment with the Texture Analyzer. To create a model of firmness variation over time, a group of tomatoes of the same variety and origin were measured for firmness at various storage dates following harvest in subsequent experiments. An experiment was conducted to measure and predict the variation in the firmness of cherry tomatoes using a tactile sensor. The results indicate that the firmness of cherry tomatoes decreases linearly over time under similar conditions. The relative prediction errors were found to be mostly within ± 15% and ± 25% during 1 and 2 weeks of storage, respectively. All findings indicate that the proposed method can effectively measure the firmness of cherry tomatoes and accurately predict their short-term variations.