Detection of early bruise in apple using near-infrared camera imaging technology combined with deep learning

Early bruises in apples, especially those occurred within 30 min, resemble healthy tissue and are difficult to be distinguished with the naked eye, making conventional techniques such as manual and machine vision sorting difficult. Near-infrared (NIR) camera has strongly ability to show bruises as bruised tissue is more sensitive to NIR light. In this paper, the NIR camera imaging technology combined with deep learning method to detect early bruise in apple was proposed. A NIR camera in the spectral range of 900-2350 nm was used to acquire early apple bruise images. After datasets of bruise images with different impact forces and different bruise duration time were established, three target detection algorithms, Faster R-CNN, Yolov3-Tiny and Yolov5s, were used to extract apple bruise area for damage detection. The results showed that the accuracy of three algorithms for early bruised and no bruised apples were all more than 99%, the accuracy of no bruised, mildly bruised and severely bruised apples were all more than 96%, and the shortest detection speed of a single image was 6.8 ms. This study shows that this method has high detection accuracy and fast recognition speed. It can be effectively applied to early bruise detection in apples, and is expected to realize real-time online detection.