A real-time object detection method of pineapple ripeness based on improved YOLOv8 was proposed to improve the accuracy of mechanical harvesting of pineapples in planting areas with different ripeness and ensure the quality of pineapples. The improved model replaced the common convolutions in the backbone and neck parts of the original YOLOv8 model with depth-wise separable convolutions （DSConv） to streamline parameters of model to solve the problems of small object size， dense quantity， and light occlusion of picked mechanical pineapple picking in natural environments. Convolutional block attention mechanism （CBAM） module was introduced before feature fusion to prioritize important features and improve the accuracy of object detection. The original loss function CIoU of YOLOv8 network was replaced with the EIoU loss function to accelerate the speed of network convergence. The results showed that the mean of average precision （P_mA） of the improved model for detecting the pineapple ripeness was 97.33%. The P_mA of improved model was 5.53， 7.91， 4.38， and 4.66 percentage points higher than that of Faster R-CNN， YOLOv4， YOLOv5 and YOLOv7， respectively. The number of parameters of the algorithm model was only 16.8×10⁶ on the premise of ensuring the accuracy of detection. It is indicated that the improved model improves the accuracy and inference speed of recognizing pineapple ripeness， and has stronger robustness.