We present a pipeline for the visual localization and classification of agricultural pest insects by computing a saliency map and applying deep convolutional neural network (DCNN) learning. First, we used a global contrast region-based approach to compute a saliency map for localizing pest insect objects. Bounding squares containing targets were then extracted, resized to a fixed size and used to construct a large standard database calledPest ID. This database was then utilized for self-learning of local image features which were, in turn, used for classification by DCNN. DCNN learning optimized the critical parameters, including size, number and convolutional stride of local receptive fields, dropout ratio and the final loss function. To demonstrate the practical utility of using DCNN, we explored different architectures by shrinking depth and width and found effective sizes that can act as alternatives for practical applications. On the test set of paddy field images, our architectures achieved a mean Accuracy Precision (mAP) of 0.951, a significant improvement over previous methods.

Background: Convolvulus sepium (hedge bindweed) detection in sugar beet fields remains a challenging problem due to variation in appearance of plants, illumination changes, foliage occlusions, and different growth stages under field conditions. Current approaches for weed and crop recognition, segmentation and detection rely predominantly on conventional machine-learning techniques that require a large set of hand-crafted features for modelling. These might fail to generalize over different fields and environments. Results: Here, we present an approach that develops a deep convolutional neural network (CNN) based on the tiny YOLOv3 architecture for C. sepium and sugar beet detection. We generated 2271 synthetic images, before combining these images with 452 field images to train the developed model. YOLO anchor box sizes were calculated from the training dataset using a k-means clustering approach. The resulting model was tested on 100 field images, showing that the combination of synthetic and original field images to train the developed model could improve the mean average precision (mAP) metric from 0.751 to 0.829 compared to using collected field images alone. We also compared the performance of the developed model with the YOLOv3 and Tiny YOLO models. The developed model achieved a better trade-off between accuracy and speed. Specifically, the average precisions (APs@IoU0.5) of C. sepium and sugar beet were 0.761 and 0.897 respectively with 6.48 ms inference time per image (800 × 1200) on a NVIDIA Titan X GPU environment. Conclusion: The developed model has the potential to be deployed on an embedded mobile platform like the Jetson TX for online weed detection and management due to its high-speed inference. It is recommendable to use synthetic images and empirical field images together in training stage to improve the performance of models.