Object detection in dynamic environmental conditions using evolutionary multimodal approach

Publication Type:
Issue Date:
Full metadata record
Environmental dynamism and uncertainty can play a critical role in many problems involving camera-based detection of real-life objects. Uncertainty is witnessed due to the presence of climatic irregularities including illumination changes, smoke, heat-waves, dust and rain. In such scenarios, the visibility of an object can severely be influenced by both signal-noise and occlusion. With the recent developments in sensing technology and computing domains, it is still possible to overcome the shortcomings of uncertainty. Multimodal image processing techniques provide very encouraging results by reducing noise and improving visibility. However, the multimodality needs further improvements to enhance accuracy, performance and robustness. Here, an evolutionary multimodal method is proposed to succeed over the discussed limitations. An evolutionary biological inspiration is applied to create a set of computing models. The proposed set of innovative evolutionary algorithms allows to reduce redundancies in datasets and improve the detection process. Experimental validation is performed for testing proposed algorithms. A formal simulation method for data modelling process was incorporated in the testing scheme to emulate environmental variations. Rigorous experimenting and analysis show the merits of the proposed methodology. Notably, both the accuracy and performance can be improved significantly using the proposed evolutionary apparatus.
Please use this identifier to cite or link to this item: